educação musical, percepção musical e suas relações com a leitura
TRANSCRIPT
Hugo Cogo-Moreira
EDUCAÇÃO MUSICAL, PERCEPÇÃO MUSICAL E SUAS RELAÇÕES
COM A LEITURA DE CRIANÇAS COM PROBLEMAS DE LEITURA:
UMA REVISÃO SISTEMÁTICA, ENSAIO CLÍNICO RANDOMIZADO SEM
PLACEBO E MODELAGEM ESTRUTURAL
Tese apresentada à Universidade Federal de
São Paulo – Escola Paulista de Medicina,
para obtenção do Título de Doutor em
Ciências.
São Paulo
2012
Hugo Cogo-Moreira
EDUCAÇÃO MUSICAL, PERCEPÇÃO MUSICAL E SUAS RELAÇÕES
COM A LEITURA DE CRIANÇAS COM PROBLEMAS DE LEITURA:
UMA REVISÃO SISTEMÁTICA, ENSAIO CLÍNICO RANDOMIZADO SEM
PLACEBO E MODELAGEM ESTRUTURAL
Tese apresentada à Universidade Federal de
São Paulo – Escola Paulista de Medicina,
para obtenção do Título de Doutor em
Ciências.
Orientador:
Prof. Dr. Jair de Jesus Mari
Coorientadora:
Profa. Dra. Clara Regina Brandão de Ávila
São Paulo
2012
Cogo-Moreira, Hugo
Educaçã musical, percepção musical e suas relações com a leitura de
crianças com problemas de leitura: uma revisão, ensaio clínico
randomizado sem placebo e modelagem estrutural / Hugo Cogo-Moreira.
– São Paulo, 2012.
xiii, 130f.
Tese (Doutorado) – Universidade Federal de São Paulo. Escola Paulista
de Medicina. Programa de Pós-Graduação em Psiquiatria e Psicologia
Médica.
Título em inglês: Music education, musical perception and their relation
with reading among poor reader school-children: a review, a randomizaed
open-label clinical trial and structural equation modeling.
1. Dislexia. 2. Musica. 3. Percepção Auditiva. 4. Ensaio Clínico
Controlado Aleatório.
iii
UNIVERSIDADE FEDERAL DE SÃO PAULO
ESCOLA PAULISTA DE MEDICINA
DEPARTAMENTO DE PSIQUIATRIA
Chefe do Departamento:
Profa. Dra. Julieta Freitas Ramalho da Silva
Coordenador do Curso de Pós-graduação:
Prof. Dr. Jair de Jesus Mari
iv
Hugo Cogo-Moreira
EDUCAÇÃO MUSICAL, PERCEPÇÃO MUSICAL E SUAS RELAÇÕES
COM A LEITURA DE CRIANÇAS COM PROBLEMAS DE LEITURA:
UMA REVISÃO SISTEMÁTICA, ENSAIO CLÍNICO RANDOMIZADO SEM
PLACEBO E MODELAGEM ESTRUTURAL
Presidente da banca:
Jair de Jesus Mari
BANCA EXAMINADORA
Ângela Maria Viera Pinheiro
Evandro da Silva Freire Coutinho
Guilherme Vanoni Polanczyk
Lia Vera Tomás
Suplentes:
Christian Costa Kieling
Elisa Brietzke
v
Dedicatória
Aos meus queridos pais, Ozeas e Lourdes,
que sempre apoiaram as minhas escolhas, seja como um
pianista nada promissor ou com um cientista, sendo sempre os
primeiros a acreditar em meu sonhos – um deles aqui escrito.
vi
Agradecimentos
Primeiramente, agradeço a Profa. Latife Yazigi que me abriu as portas do
Departamento de Psiquiatria da UNIFESP, acolhendo-me após um tempestuoso
inverno no Departamento de Ortopedia da mesma instituição. Sem ela, todas as
próximas as palavras a partir daqui tecidas não existiriam.
Ao Prof. Jair de Jesus Mari que assumiu a minha orientação após um ano sobre os
cuidados da Profa. Latife Yazigi confiando plenamente em meu trabalho e idéias
megalomaníacas. Sem esta plena confiança, apoio irrestrito e liberdade para criar,
todos os estudos, financiamentos e descobertas apresentados aqui não existiriam.
À Profa. Clara Brandão de Ávila do Departamento de Fonoaudiologia da UNIFESP
que, carinhosamente e de pronto, acolheu a mim e meu estudo em um momento que
eu tinha acabo de ser apresentado a área de investigação científica sobre dificuldades
de leitura.
À Profa. Ângela Pinheiro que me cedeu suas listas inéditas de palavras e não-
palavras e a Escala de Avaliação da Competência Leitora pelo Professor (EACOL) –
instrumentos sem os quais esse estudo não poderia ter sido realizado de forma
validada e fidedigna.
Às minhas amigas Carolina Carvalho, Ghina (Re)Ghina Dourado Machado e Nádia
Shigaeff que durante o trabalho de campo foram os meus braços, pernas e
conselheiras na organização de um time de 30 profissionais de saúde (entre psicólogos
e fonoaudiólogos) na avaliação de mais de 900 crianças em dez escolas públicas na
cidade de São Paulo.
Ao Instituto Rukha e aos Parceiros da Educação (um especial agradecimento a
Mônica Guerra e cada um dos facilitadores) que me apresentaram nove das dez
escolas que participaram desse estudo, acolhendo o meu projeto e idéias.
vii
As dez escolas, seus diretores e professores que carinhosamente acolheram minha
idéia, os fonoaudiólogos e psicólogos durante a execusação doprojeto, mostrando que
SIM existem pessoas preocupadas com a educação e desenvolvimento intelectual de
nosso País, fazendo mais do que seu papel como educadores, sendo vezes pais/mães.
Agradeço os sorrisos, carinho e todos os votos de confiança depositados em mim e
minha equipe. Como uma das escolas, a Gonzaquinha, não foi apresentada nem pelo
instituto Rukha nem pelo Parceiros da Educação, gostaria de fazer um agradecimento
especial, já que me receberam o projeto igualmente e prontamente.
Aos professores de música que participaram do ensaio clínico e que muito me
ensinaram sobre educação musical.
As crianças que participaram bravamente desse estudo que teve em seu total mais de
3 horas entre avaliações fonoaudiológicas, comportamentais, de inteligência e de
habilidades musicais (esse ultima avaliação, diga-se de passagem, extremamente
chata e cansativa).
À amiga do mundo acadêmico de longa data que me apoiou incondicionalmente
nesse trabalho hercúleo Profa. Lia Vera Tomás.
Aos bons amigos que fiz nesse percurso UNIFESP[iano] pelas palavras de afeto e
motivação Hudson Carvalho, Profa. Denise Razzouk, Taís Moryama, Giovanni
Salum Jr. e Luciana Moura – mesmo à distância de mais de 9.000 quilômetros.
A Denise Sessa, Cláudia e Fábia que me ensinaram a como administrar dinheiro
público.
Ao Prof. George Ploubidis e sua equipe que me acolheu na London School of
Hygiene and Tropical Medicine durante o estágio de doutorado sanduíche.
Ao Instituto ABCD e ao CNPq (Edital Universal) pela concessão do financiamento que
permitiu a realização desse estudo. À CAPES, tanto pela bolsa de estudo no Pais
quanto a bolsa de estudos no exterior, e a AUTOMAM Engenharia pelo complemento
a bolsa de estudo no Pais quanto a bolsa de estudos no exterior.
viii
Sumário
Suplentes: ................................................................................................................................. iv
Dedicatória ................................................................................................................................. v
Agradecimentos ........................................................................................................................ vi
Lista de figuras .......................................................................................................................... ix
Lista de tabelas .......................................................................................................................... x
Lista de abreviaturas ................................................................................................................. xi
Resumo ..................................................................................................................................... xii
Abstract .................................................................................................................................... xiv
1. INTRODUÇÃO ...................................................................................................................... 1
1.1 Apresentação desta Tese ................................................................................................ 2
1.2 Quais são as evidências, até o presente momento, da efetividade da educação musical para o aprimoramento das habilidades de leitura? .......................................................... 2
1.3 O primeiro ensaio clínico randomizado da literatura: qual a efetividade da educação musical sobre habilidades de leitura e desempenho acadêmico em crianças com dificuldades de leitura? .................................................................................................... 4
1.4 Como mensurar fidedignamente leitura de escolares e identificar aqueles com problemas de leitura? Quais são os instrumentos disponíveis em Português do Brasil? . 6
1.5 Como se dariam as interações entre habilidades de percepção musical e leitura? ......... 9
2. REFERÊNCIAS ................................................................................................................... 11
3. ARTIGOS ............................................................................................................................ 15
3.1 Music education for improving reading skills in children and adolescents with dyslexia (Review) ........................................................................................................................ 16
3.2 EACOL (Scale of Evaluation of Reading Competency by the Teacher): Evidence of Concurrent and Discriminant Validity ............................................................................. 42
3.3 Effectiveness of Music Education for the Improvement of Reading Skills and Academic Achievement in Young Poor Readers: a Pragmatic Cluster-Randomized, Controlled Clinical Trial ................................................................................................................... 70
3.4 Music Perception Predicts Word-level Reading Ability in Children with Reading Difficulties ...................................................................................................................... 97
ix
Lista de figuras
Artigo 1
Figura 1 – Study flow Diagram...........................................................................................
24
Artigo 2
Figura 1 – Latent Classes for BH-sample.................................................................... 54
Figura 2 – Latent Classes for SP-screening................................................................. 55
Figura 3 – Latent Classes for SP-trial……………………………………………………..
56
Artigo 4
Figura 1 – Integrative model with convergence problems.................................................. 109
Figura 2 – Integrative model with standardised coefficients............................................... 110
x
Lista de tabelas
Artigo 2
Tabela 1 – Latent Classes Analysis Results……………………………………………. 57
Tabela 2 – Values for Regression Coefficients with its respective robust standard error, p-value and 95% confidence interval for variables of Concurrent
and Discriminant Validity ……………………………………......................
60
Artigo 3
Tabela 1 – Measurements and them comparison between control and intervention schools..................................................................................................
94
Tabela 2 – Effects of Music Education considering ITT and CACE.............................. 95
Tabela 3 – Intraclass correlation coefficient for primary and secondary outcomes at baseline and last assessment....................................................................
96
Artigo 4
Tabela 1 - Mean, standard deviation (SD), kurtosis, skewness, and univariate normality test (skewness and kurtosis normality test) for all continuous (observable and latent) variables...................................................................
107
xi
Lista de abreviaturas
AIC Akaike Information Criterion
BIC Bayesian Information Criterion
CACE Complier-average Causal Effect
CFA Confirmatory Factor Analysis
CI Confidence Interval
EACOL Escala de Avaliação da Competência Leitora pelo Professor
ICC Intraclass coefficient correlation
IQ Intelligence Quotient
ITT Intention-to-treat
LCA Latent Class Analysis
NPC National Parameters Curriculum
RA Reading Aloud
RSE Robust Standard Error
SDQ Strength and Difficulties Questionnaire
SR Silent Reading
ssaBIC Sample size adjusted Bayesian Information Criterion
xii
Resumo
Objetivos: Esta tese investigou evidências da efetividade da educação musical e de
como as habilidades de percepção musical poderiam colaboram para o desempenho
de habilidades de leitura para crianças com problemas de leitura, por meio de quatro
estudos com diferentes desenhos metodológicos: 1) uma revisão sistemática da
literatura registrada no Cochrane Developmental, Psychosocial, and Learning Problems
Groups, a qual investigou os efeitos do aprender musical para a melhora de habilidades
de leitura entre crianças e adolescentes com dislexia; 2) um estudo de validação de
uma escala de avalição de habilidades leitoras aplicada pelo professor; 3) um ensaio
clínico randomizado em bloco que visou verificar a efetividade de um um programa de
intervenção musical, em período extracurricular, baseado nos Parâmetros Curriculares
Nacionais, três vezes por semana durante 5 meses, a uma população de 235
escolares de oito a dez anos de idade com dificuldades de leitura distribuídos em 10
escolas públicas de regiões periféricas da cidade de São Paulo; 4) um modelo teórico
integrativo, via análise de trajetória com os dados do baseline do ensaio clínico
randomizado, que vislumbrou compreender como se dão os mecanismos subjacentes a
possíveis melhoras em habilidades de leitura (circunscritas as habilidades no nível da
decodificação de palavras isoladas e em contexto de leitura avaliada pelo Montreal
Battery of Evaluation of Amusia). Método: para a revisão sistemática, uma estratégia
sensível foi criada e apartir dos resultados dois dos autores selecionaram os títulos
Métodos: a revisão sistemática foi teve uma estratégia de música muito sensível,
lançadas não somente em bases de dados na área da medicina e psicologia, mas
também em bases de dados das ciências sociais e artes (dada a natureza da
intervenção). Não houve restrição de idiomas. Os resultado da busca (876 títulos)
foram avaliados por dois dos autores do trabalho independentemente, sendo alto o
nível de concordância entre eles. Para o estudo de validação da EACOL (Escala de
Avaliação da Compreensão Leitora), resultados obtidos por meio de Análise de Classe
latentes foram usados em modelos de regressão para avaliar a validade discriminante
e concorrente do constructo subjacente a tal escala.
No que tange o ensaio clínico, os dados das 235 crianças foram analisados sobre dois
paradigmas considerando o desenho em multiníveis: uma tradicional, a itenção de
tratar, e a outra chamada de CACE – Complier-average causal effect – um método de
estimação subjacente aos modelos mistos da modelagem de equações estruturais que
xiii
permite estimar efeitos para aquelas crianças que aderiram e não aderiram as aulas de
música. Os modelo teórico integrativo foi feito por meio de modelagem de equação
estrutural. Resultados: apesar do senso comum de que a educação musical seja
benéfica para problemas de leitura, nenhum ensaio clínico randomizado foi encontrado
dentre as 876 citações resultantes da estratégia de busca. No que tange a escala, os
27 itens que a constituem a EACOL mostraram ser discriminates, sendo capazes de
categorizar de forma fidedigna três tipologias de leitores.
Quanto ao ensaio clínico, os escolares que receberam intervenção liam ao final do
estudo 2,57 (p-value= 0,047) palavras corretas lidas por minuto a mais que o grupo
controle, desempenho escolar português de 0,21 (p-value <0,001) a mais na média a
cada bimestre e 0.246 (p-value <0.001) a mais para o desempenho em matemática. Ou
seja, ou final de quatro bimestres, a diferença entre os grupos era de 0.84 para
português e 0.96 em matemática. Considerando a análise por adesão, CACE, (mínimo
de 1% de presença às aulas de música), as estimativas são superiores às descritas.
Para algumas medidas o efeito da intervenção mostrou-se quase três vezes superior
(notas em Português; 0,77 por bimestre), em matemática quase meio ponto por
bimestre (β = 0,491, p-value<0,001) e a leitura de palavras isoladas teve um índice seis
vezes maior (β = 13,98, p-value <0.001). Quanto ao modelo teórico, um fator geral
latente do modelo de percepção musical e uma parte específica do mesmo (o domínio
melódico) mostraram-se preditores do desenvolvimento da leitura no nível da palavra.
Conclusões: Apesar dos achados promissores do primeiro ensaio clínico no mundo
dos efeitos de aprender música em habilidades de leitura e desempenho acadêmico
em jovens escolares, ainda seria muito prematuro em dizer que temos a panacéa dos
problemas educaionais, mas os resultados são animadores do ponto de vista
educacional quanto da neurociências. Ainda, como resultados dessa tese, apresenta-
se a EACOL como um instrumento de fácil aplicação, rápido e com bom ajustamento
teórico e estatístico para avaliar a competência leitora de jovens escolares por meio do
professor.
xiv
Abstract
Aims: This thesis investigated evidence of the effectiveness of music education for
children with reading difficulties through a systematic literature review and a randomized
clinical trial, without placebo, about the effects of an extracurricular music intervention in
a population of schoolchildren from eight to ten years old with reading difficulties. In
addition, to understand better the mechanisms underlying the possible improvements, a
theoretical model which integrated music perception abilities and reading skills was
designed and tested. This thesis also validated a scale developed for the evaluation of
reading competency by the teacher for children from the second to fourth years of
primary school. Methods: For the systematic review study, two authors independently
screened all titles and abstracts identified through search strategy to determine their
eligibility. In order to validate EACOL, results from a latent class analysis LCA, in which
two latent groups were obtained as solutions, and were correlated with direct and
indirect reading measures, providing concurrent and disciminant validity. Regarding to
evaluation of effectiveness of music education for reading and academic skills
improvement, 235 children with reading difficulties in 10 schools participated in a five-
month, randomized clinical trial in cluster (RCT) in an impoverished zone within the city
of São Paulo to test the effects of music education intervention while assessing reading
skills and academic achievement during the school year. Five schools were chosen
randomly to incorporate music classes (n = 114), and five served as controls (n = 121).
Two different methods of analysis were used to evaluate the effectiveness of the
intervention: the standard method was intention-to-treat (ITT), and the other was the
Complier Average Causal Effect (CACE) estimation method, which took compliance
status into account. Lastly, in order to explain how music perception skills may act as
predictos of word-level reading skills, a general-specific solution of Montreal Battery of
Evaluation of Amusia (MBEA), which underlies a music perception construct and is
constituted by three latent factors (the general, temporal and the melodic domain), was
regressed on word-level reading skills (rate of correct isolated words/non-words read
per minute).
Results: Despite the common view that music education is beneficial to students with
reading problems, no randomized trial has been found among the 876 citations which
xv
resulted from a sensitive search strategy without language restriction. As for the clinical
trial, children who received intervention showed trends and indexes statistically
significant for improvement in the outcomes of reading and academic performance, as
measured by the grades in Portuguese and mathematics subjects during a school year.
Considering the analysis by adherence (at least 1% of presence in music lessons), the
estimates are higher than those observed. For some measures, the effect of the
intervention proved to be almost three times as great (notes in Portuguese).
Regarding the theoretical model, musical perception as part of basic auditory perception
may contribute to the development of word-level reading, especially in the case of non-
word reading and phonological awareness. A recursive model was generated,
confirming that it was also a mediator of fluency and a predictor of word-level reading.
Last, the Scale of Evaluation of Reading Competency by the Teacher (EACOL) is a
useful screening tool for assessing reading skills and for predicting three different
typologies of 2nd-4th grade readers.
1 . I N T R O D U Ç Ã O
I n t r o d u ç ã o | 2
1.1 Apresentação desta Tese
Na introdução desta tese, apresentam-se as quatro perguntas norteadoras das
investigações que desencadearam a produção dos quatro artigos e suas respectivas
conclusões. Embora cada artigo se estruture segundo pressupostos, método e,
referências diferentes, já que versam sobre questionamentos e objetos diferentes,
todos convergem com contribuições para uma melhor compreensão dos problemas de
leitura que afetam os escolares brasileiros desde os primeiros anos do ensino
fundamental. Mais adiante, serão expostas tanto os principais obstáculos para
desenvolver pesquisas na área de dificuldades de aprendizagem quanto as solução
soluções tomadas para sanar os empecilhos, da forma como se apresentavam. As
investigações realizadas para responder às quatro questões resultaram em quatro
artigos científicos, aqui descritos como estudos.
1.2 Quais são as evidências, até o presente momento, da efetividade da
educação musical para o aprimoramento das habilidades de leitura?
Associações entre habilidades de percepção e prática musical e linguagem são
foco de investigação científica em diferentes áreas do conhecimento, não se
restringindo exclusivamente à música ou à linguística. Psicologia, neurologia e acústica
são áreas que têm se voltado para o desenvolvimento de modelos teóricos e
experimentais nesse campo interdisciplinar que envolve as ciências da saúde e a
música, considerado como inovação para o melhor entendimento sobre plasticidade
cerebral.(1) Nesse contexto, o seguinte questionamento surgiu: quais são as evidências
a respeito dos efeitos e impacto do aprendizado musical sobre problemas de leitura,
em especial em crianças com dislexia nas quais as habilidades de leitura
(decodificação, compreensão textual) são comprometidas?
O aprender musical tem sido associado a benefícios como, por exemplo,
aprimoramento de inteligência verbal e função executiva,(2) aprimoramento de
habilidades linguísticas(3-6) e habilidades motoras finas.(7, 8) Esses exemplos são apenas
uma pequena parcela das grandes possiblidades experimentais em que o aprender
musical tem sido explorado.
I n t r o d u ç ã o | 3
Um dos argumentos que suportam a existência de relação entre o aprender
musical com o aprimoramento de habilidades de leitura, baseia-se na similaridade
dessa aprendizagem com a de um novo sistema de decodificação necessário para se
aprender a tocar um instrumento, levando-se em consideração que enquanto uma é
linguagem verbal - escrita, a outra, não verbal – música.(9) Assim, é importante ressaltar
que a notação musical é a representação espacial dos padrões de alturas e duração do
som. Em contraste, o alfabeto é uma representação arbitrária dos padrões fonotáticos
de uma língua. Contudo, evidências sobre relações entre a educação musical e a
melhora de desfecho vinculada às habilidades de leitura não são conclusivas, tanto do
ponto de vista experimental quanto do teórico.
Investigar quais eram as evidências do aprimoramento de habilidades de leitura
em crianças com problemas de leitura por meio da educação musical foi a primeira
pergunta feita e que norteou a condução do primeiro artigo. Para responder tal questão,
inicialmente, desenvolvemos uma revisão sistemática a respeito das evidências da
efetividade da educação musical para o aprimoramento da leitura em crianças e
adolescentes disléxicos. Nessa revisão apresentaram-se definições sobre a dislexia,
suas causas, fatores preditores e comorbidades. Ainda, buscou-se definir a educação
musical, por meio de considerações que transcenderiam o tradicional processo de
ensino de partituras e o domínio técnico instrumental como um fim. Pode parecer
completamente familiar, e óbvio, o que seria aprender música, mas o esclarecimento a
respeito das possibilidades de abrangência do processo educacional, para além da
decodificação de partituras e da maestria com que se toca instrumentos musicais
tradicionais, como o piano e violinos, envolve diretamente o próprio conceito de música.
Portanto, exploraram-se as diferentes abordagens de aprendizado musical, tanto do
ponto de vista da fundamentação teórica quanto a prática propriamente dita.
A realização da revisão sistemática foi importante passo para conhecer as reais
evidências construídas até os dias de hoje delineando o Estado da Arte sobre o que se
conhece da efetividade do aprendizado musical para dislexia. Ainda, nessa revisão,
apresentam-se as principais teorias sobre o motivo do aprender musical ser encarado
como uma possibilidade de intervenção para problemas de leitura.
A estratégia de busca na revisão sistemática foi extremamente sensível tanto
para a questão terminológica relacionada à dislexia como, por exemplo, reading
I n t r o d u ç ã o | 4
difficulties, reading disabilities, reading problems quanto para a da educação musical,
em que termos como Kodaly method, Suzuki Method, music, timbre, foram explodidos
para se abarcar o maior número de estudos possíveis já realizados. Na estratégia de
busca, não se limitou o idioma dos estudos.
A revisão sistemática da literatura não gerou metanálise, pois, admiravelmente,
dentre as 851 citações (entre artigos, teses, dissertações, protocolos de estudo, cartas
ao editor), de diferentes bases de dados, nenhum ensaio clínico randomizado, acerca
dos efeitos da educação musical para crianças e adolescentes disléxicos, foi
encontrado.
Ressalta-se que se buscou por estudos cujos desenhos epidemiológicos fossem
ensaios clínicos randomizados (e suas variantes) já que esse é o melhor desenho para
se verificar a causalidade entre uma dada intervenção para um dado desfecho.
1.3 O primeiro ensaio clínico randomizado da literatura: qual a efetividade da
educação musical sobre habilidades de leitura e desempenho acadêmico em
crianças com dificuldades de leitura?
A falta de evidências a respeito da efetividade da educação musical sobre
habilidades de leitura e desempenho acadêmicos em crianças com problemas de
leitura nos motivou a desenhar, então, o primeiro ensaio clínico da literatura para testar
os efeitos da educação musical sobre o aprimoramento das habilidades de leitura e
desempenho acadêmico (este, como um desfecho secundário) para crianças com
dificuldades de leitura.
Nesse ponto, optamos por trabalhar com dificuldades de leitura de modo geral,
já que dislexia necessita de equipe interdisciplinar para obtenção do diagnóstico. Não
existe um teste específico para identificação dessa patologia, a qual é descrita pelo
DSM-IV como um transtorno específico de aprendizagem, caracterizado por
desempenho escolar em leitura/escrita inferior ao esperado para a idade cronológica,
escolaridade e nível cognitivo/intelectual do indivíduo.(10)Portanto, em virtude da
dificuldade envolvendo o diagnóstico, descartamos a possibilidade de um ensaio clínico
envolvendo tal transtorno específico de aprendizagem.
I n t r o d u ç ã o | 5
Ser disléxico (transtorno específico de leitura – specific reading
disorder/disability) implica possuir dificuldades de leitura com base neurobiológica,
sendo um transtorno altamente hereditário que afeta de 5 a 10% da população infantil
em fase escolar.(11, 12) Por outro lado, dificuldades de leitura podem emergir da dislexia
ou ainda de fatores relacionados com antecedentes desenvolvimentais como
estimulação em casa, fatores sócio econômicos, depressão maternal e negligência
infantil.(13)
No artigo do ensaio clínico randomizado para o aprimoramento das habilidades
de leitura e desempenho acadêmico (nas disciplinas de português e matemática no
decorrer do ano letivo) apresenta-se a descrição de todo o processo de concepção do
ensaio clínico: a seleção das escolas, o cálculo amostral, a randomização das escolas
e explicação da intervenção educacional musical que foi adotada. Ainda, retomou-se a
discussão sobre o que seria o processo de aprendizagem musical, tendo como
referência a estrutura dos Parâmetros Curriculares Nacionais (PCN), desenvolvido na
década de 70 e que está em consonância com os modernos métodos de aprendizado
musical.
Na introdução desse que, na ordem de apresentação, será o terceiro artigo,
enfatizaram-se as evidências neurobiológicas a respeito de como o aprendizado
musical poderia ajudar no desenvolvimento da leitura.(14, 15)
Indiferentemente ao fato do objeto de estudo de um ensaio clínico ser dislexia ou
dificuldades de leitura, uma importante limitação da área dos estudos sobre a leitura de
escolares falantes de língua portuguesa, é a ausência de instrumentos validados com
bons indicadores psicométricos para avaliar as habilidades de leitura. Em virtude desse
grande limitador para essa área de estudo decidiu-se, antes de tudo, validar uma
escala de avaliação leitora que tem como parâmetro uma medida indireta da análise da
leitura – a observação do professor. A Escala de Avaliação da Competência Leitora
pelo Professor (EACOL) foi desenvolvida pela professora Ângela Pinheiro da
Universidade Federal de Minas Gerais em parceira com a professora Ana Edith. É
constituída por 27 itens dicotômicos, sendo que 17 deles investigam a leitura em voz
alta, e dez itens, a leitura silenciosa de escolares de terceiro ao quinto ano do ensino
fundamental. O artigo da validação da EACOL é apresentado aqui como o segundo
artigo.
I n t r o d u ç ã o | 6
1.4 Como mensurar fidedignamente leitura de escolares e identificar aqueles
com problemas de leitura? Quais são os instrumentos disponíveis em
Português do Brasil?
Pouco são os instrumentos disponíveis no mercado para avaliar diferentes
domínios da leitura em Português do Brasil de forma sistemática que passaram por
validação, normatização e análise de constructo. Dentre alguns podemos citar o Teste
do Desempenho Escolar (TDE)(16) destinado à avaliação dos níveis de escrita, leitura e
aritmética nas quatro primeiras séries do ensino fundamental. É um instrumento
padronizado, mas com algumas limitações do ponto de vista do constructo teórico e no
que se diz respeito à sua normatização.
No domínio da consciência fonológica, habilidade de manipular, mentalmente, os
sons que constituem as palavras, cita-se o CONFIAS(17) e o Teste de Consciência
Fonológica(18) que não ainda passaram por rigorosos testes de validade de constructo
tal como análise fatorial confirmatória ou mesmo análise de classes latentes. No que
tange à estrutura dos itens/estímulos de ambos os testes, podem-se fazer algumas
considerações a respeito do poder de discriminação: alguns itens são extremamente
fáceis e outros extremamente difíceis para as crianças e, dessa forma, o poder de
discriminação do item per se fica prejudicado, sendo fundamental que a bateria ou
teste, como um todo, seja equalizada e apresente itens discriminativos. Do ponto de
vista do item em relação ao constructo, a análise verificando o domínio subjacente ao
conjunto de itens, normalmente, não é realizada, seja verificando se o constructo
subjacente é contínuo ou categórico (em classes). Por fim, ambos os testes de
consciência fonológica não passaram por normatização.
No nível da decodificação (nível de leitura de palavras e não-palavras), poucas
são as tarefas de leitura, em voz alta, disponíveis. O TDE contém uma lista de palavras
no subteste de leitura, não controlada, do ponto de vista linguístico de seus itens, para
as direções de leitura e de escrita. Ou seja, como sugerem os estudos envolvendo o
fenômeno da retroalimentação fonológica, o TDE pode ser criticável sob esse
aspecto.(19)
No que tange aos elementos que envolvem habilidades cognitivas superiores,
por exemplo, a compreensão leitora, somente recentemente as pesquisas brasileiras
I n t r o d u ç ã o | 7
começaram a propor instrumentos. Na PUC, a Professora Solange Wechsler
recentemente adaptou, traduziu e validou a bateria Woodcock-Johnson III para o
português do Brasil.(20-22) Tal bateria abrange uma série de domínios da linguagem
como a expressão oral e escrita, compreensão leitora, fluência de leitura e habilidades
matemáticas. Entretanto, será de uso restrito dos profissionais da área da psicologia tal
como a escala Wechsler de Inteligência (WISC-III)(23) para a avaliação de habilidades
cognitivas.(24, 25)
Recentemente, Professora Clara Brandão de Ávila, Departamento de
Fonoaudiologia da UNIFESP, obteve um financiamento para o desenvolvimento de
uma bateria para avaliação da compreensão leitora em escolares do ciclo I, via
FAPESP (2011/11369-0), sendo a primeira iniciativa no que tange à construção, com
tecnologia puramente nacional, de um instrumento capaz de avaliar as habilidades de
compreensão; o processo de análise inferencial dos testes e subtestes de tal bateria
abrangerão desde modelagem de equações estrutural e teoria de resposta ao item,
testando assim os constructos psicométricos do item até a estrutura latente superior – a
compreensão em si. O projeto se encontra em execução no momento.
Portanto, de modo geral, pode-se perceber que a carência de instrumentos de
avaliação é um limitador para a realização de estudos em que o desfecho, seja ele
primário ou secundário, é o aprimoramento de habilidades de leitura, indiferentemente
de ser um ensaio clínico randomizado ou qualquer outro desenho epidemiológico ou da
patologia ou dificuldade sobre a qual se quer intervir. Além disso, mais do que sermos
capazes de avaliar possíveis benefícios de intervenções, uma realidade que nos aflige
mediante tal carência é a falta de critérios claros, sensíveis e específicos para triar e
identificar crianças com dificuldades de leitoras. Portanto, como iniciar um diagnóstico
ou um ensaio clínico a respeito de efeitos do aprender musical sobre habilidades de
leitura quando não conseguimos mensurá-las com indicadores fidedignos e
reconhecidos? Como reconhecer crianças com dificuldades de leitura?
Uma possibilidade seria utilizar o crivo do professor acreditando, ser este,
suficientemente específico e sensível para detectar os problemas de leitura entre os
escolares. Entretanto, surge aqui uma importante discussão na literatura internacional,
retomada em 2011 após 20 anos, sobre a capacidade de identificação do professor em
discriminar as habilidades de leitura de seu alunado.
I n t r o d u ç ã o | 8
Surge, então, a motivação para o segundo artigo, que explora, logo na
introdução, esse “gap” da literatura e apresenta as principais evidências a respeito da
acurácia do professor como uma medida indireta das habilidades de leitura dos
escolares. Como objetivo central, o estudo propôs validar uma escala desenvolvida
pela Professora Ângela Maria Vieira Pinheiro da Universidade Federal de Minas Gerais
chamada de EACOL – Escala de Avaliação da Competência Leitora pelo Professor.
Para tal, várias medidas diretamente relacionadas com leitura (p.ex., número de
palavras lidas corretamente por minuto, número de não-palavras lidas corretamente por
minuto, número de palavras corretas lidas em um texto) foram usadas como medidas
para validação convergente, ou seja, verificando se o constructo subjacente à EACOL,
de fato, mensura habilidades de leitura. Para a validação divergente, ou seja, verificar
se o constructo subjacente à EACOL não está relacionado com medidas não
vinculadas com o constructo de leitura, por exemplo, foram usados sintomas
psiquiátricos, mensurados via questionário de Capacidades e Dificuldades – Strengths
and Difficulties Questionnaire (SDQ)(26, 27) e a escala Wechsler de inteligência,(25) ambos
os instrumentos validados, traduzidos e adaptados para o Português do Brasil.
A EACOL, com seus 27 itens dicotômicos se mostrou válida para mensurar
habilidades de leitura e ainda permitiu a discriminação de três grupos de leitores, tal
como previsto teoricamente pela autora.
Essa escala permitiu, então, distinguir diferentes grupos de leitores e, ao mesmo
tempo e indiretamente, validar as medidas diretas de habilidades de leitura no nível da
palavra.
Dado que temos um instrumento capaz de distinguir grupos de leitores e
medidas diretas de leitura validadas, foi possível desenvolver o primeiro ensaio clínico
a respeito da efetividade da educação musical para o aprimoramento das habilidades
de leitura e também o desempenho acadêmico em Português e Matemática de
crianças com dificuldades de leitura que culminou com os achados do terceiro artigo.
I n t r o d u ç ã o | 9
1.5 Como se dariam as interações entre habilidades de percepção musical e
leitura?
Os resultados e tendências positivas encontrados no primeiro ensaio clínico a
respeito da efetividade da educação musical para a leitura e desempenho acadêmico
de escolares com dificuldades de leitura geraram um questionamento mais importante,
do que vislumbrar o os efeitos obtidos: como tais efeitos poderiam ocorrer, em um
campo teórico, para além das evidências do ponto de vista neuroanatômico em que se
observa que o cérebro de músicos é mais proficiente em performance de tarefas
relacionadas com a linguagem.(28-30)
Empiricamente, os principais argumentos que nutrem essa associação provêm
de analogias entre o aprendizado musical e o aprendizado verbal. O processo de
aprender música por envolver a decodificação de micro-estruturas (notas e pausas)
que se organizam para se formar elementos superiores (melodias) podem ser, de
forma análoga, associados ao processo de desenvolvimento da linguagem verbal em
que pequenas palavras formam estruturas superiores como frases.(9) Mas, por outro
lado, é importante ressaltar que o alfabeto é uma representação arbitrária dos padrões
fonotáticos de uma língua. Enquanto a partitura em si (aqui entendida como um
conjunto de notas musicais e pausas) não possui representações que permitam
estabelecimento de ordem semântica tal como acontece em textos. Nessa tese, não
será discutido qualquer processo envolvendo decodição de partituras, limitando-se
exclusivamente a questão da percepção musical, como parte da percepção auditiva.
Para melhor compreender, do ponto de vista teórico, como as habilidades de
percepção musical inatas - visto que as crianças participantes do ensaio clínico nunca
haviam passado por educação musical segundo seus pais – poderiam ser preditoras de
um melhor desempenho em tarefas de leitura de palavras e não palavras e essas, por
sua vez, poderiam melhorar a leitura de texto. Assim, saindo da esfera da
decodificação e passando para um nível hierárquico superior, desenhamos um modelo
teórico, por meio de modelagem de equações estruturais, que permitiria testar as
possíveis relações entre as estruturas subjacentes à percepção musical e à leitura de
forma integrativa. Esse foi o objetivo do quarto artigo apresentado nessa tese. Apesar
do modelo teórico ter sido desenvolvido considerando as mensurações do baseline, a
I n t r o d u ç ã o | 10
tipologia de análise empregada – a path analysis – permite verificar causalidade, mas
essa não possui a robustez do ensaio clínico randomizado.
Portanto, a organização dessa tese se apresenta da seguinte forma: 1) revisão
sistemática da literatura mundial a respeito da efetividade da educação musical para o
aprimoramento da leitura de crianças e adolescentes com dislexia; 2) validação de um
instrumento para avaliação da leitura de escolares do terceiro ao quinto ano do ensino
fundamental por meio do professor; 3) ensaio clínico randomizado sem placebo em
crianças com dificuldades de leitura; 4) modelo teórico que permitiria explicar as
relações entre as habilidades de percepção musical e as de leitura.
2 . R E F E R Ê N C I A S
R e f e r ê n c i a s | 12
1. Schlaug G, Forgeard M, Zhu L, Norton A, Winner E. Training-induced neuroplasticity in young children. Ann N Y Acad Sci. 2009 Jul;1169:205-8.
2. Moreno S, Bialystok E, Barac R, Schellenberg EG, Cepeda NJ, Chau T. Short-term music training enhances verbal intelligence and executive function. Psychol Sci. 2011 Nov;22(11):1425-33.
3. Marin MM. Effects of early musical training on musical and linguistic syntactic abilities. Ann N Y Acad Sci. 2009 Jul;1169:187-90.
4. Besson M, Schon D, Moreno S, Santos A, Magne C. Influence of musical expertise and musical training on pitch processing in music and language. Restor Neurol Neurosci. 2007;25(3-4):399-410.
5. Moreno S, Marques C, Santos A, Santos M, Castro SL, Besson M. Musical training influences linguistic abilities in 8-year-old children: more evidence for brain plasticity. Cereb Cortex. 2009 Mar;19(3):712-23.
6. Patel AD. Why would Musical Training Benefit the Neural Encoding of Speech? The OPERA Hypothesis. Front Psychol. 2011;2:142.
7. Costa-Giomi E. Does music instruction improve fine motor abilities? Ann N Y Acad Sci. 2005 Dec;1060:262-4.
8. Schneider S, Schonle PW, Altenmuller E, Munte TF. Using musical instruments to improve motor skill recovery following a stroke. J Neurol. 2007 Oct;254(10):1339-46.
9. Forgeard M, Winner E, Norton A, Schlaug G. Practicing a musical instrument in childhood is associated with enhanced verbal ability and nonverbal reasoning. PLoS One. 2008;3(10):e3566.
10. American Psychiatric Association. DSM-IV-TR: manual diagnóstico e estatístico de distúrbios mentais. Dornelles C, tradutora. 4a ed. Porto Alegre: Artes Médicas; 2002.
11. Scerri TS, Schulte-Korne G. Genetics of developmental dyslexia. Eur Child Adolesc Psychiatry. 2010 Mar;19(3):179-97.
12. Shaywitz SE, Shaywitz BA, Fletcher JM, Escobar MD. Prevalence of reading disability in boys and girls. Results of the Connecticut Longitudinal Study. JAMA. 1990 Aug 22-29;264(8):998-1002.
13. Trzesniewski KH, Moffitt TE, Caspi A, Taylor A, Maughan B. Revisiting the association between reading achievement and antisocial behavior: new evidence of an environmental explanation from a twin study. Child Dev. 2006 Jan-Feb;77(1):72-88.
14. Johansson BB. Cultural and linguistic influence on brain organization for language and possible consequences for dyslexia: a review. Ann Dyslexia. 2006 Jun;56(1):13-50.
R e f e r ê n c i a s | 13
15. Sluming V, Brooks J, Howard M, Downes JJ, Roberts N. Broca's area supports enhanced visuospatial cognition in orchestral musicians. J Neurosci. 2007 Apr 4;27(14):3799-806.
16. Stein LM. TDE: teste de desempenho escolar: manual para aplicação e interpretação. São Paulo: Casa do Psicólogo; 1994.
17. Moojen S, Lamprecht R, Santos RM, Freitas GM, Brodacz R, Siqueira M, et al. CONFIAS Consciência fonológica: instrumento de avaliação sequencial São Paulo: Casa do Psicólogo; 2003.
18. Capovilla AG, Capovilla FC. Prova de consciência fonológica: desenvolvimento de dez habilidades da pré-escola à segunda série. Temas Desenvolv. 1998;7(37):14-20.
19. Pinheiro AM. Transparência ortográfica e o efeito de retroalimentação fonológico grafêmica: implicações para a construção de provas de reconhecimento de palavras In: Alves LM, Mousinho R, Aparecida CS. Dislexia: novos temas, novas perspectivas. Rio de Janeiro: Wak; 2011. p. 131-51.
20. Wechsler SM, Vendramini CM, Schelini PW. Adaptação brasileira dos testes verbais da bateria Woodcock-Johnson III. Interam J Psychol. 2007;41(3):285-94.
21. Wechsler SM, Schelini PW. Bateria de habilidades cognitivas Woodcock-Johnson III: validade de construto. Psicol Teor Pesqui. 2006;22(3):287-96.
22. Wechsler SM, Nunes CS, Schelini PW, Pasian SR, Homsi SV, Moretti L, et al. Brazilian adaptation of the Woodcock-Johnson III Cognitive Tests. Sch Psychol Int. 2010 August 1, 2010;31(4):409-21.
23. Weschler D; Psychological Corporation. WISC-III: Weschler intelligence scale for children. 3rd ed. New York: Hartcourt Brace Jovanovich; 1991.
24. Nascimento E, Figueiredo VL. WISC-III e WAIS-III: alterações nas versões originais americanas decorrentes das adaptações para uso no Brasil. Psicol Reflex Crit, . 2002;15(3):603-12.
25. Cruz MB. WISC III: Escala de Inteligência Wechsler para crianças: manual. Aval Psicol. 2005;4(2):199-201.
26. Goodman R, Ford T, Corbin T, Meltzer H. Using the Strengths and Difficulties Questionnaire (SDQ) multi-informant algorithm to screen looked-after children for psychiatric disorders. Eur Child Adolesc Psychiatry. 2004;13 Suppl 2:II25-31.
27. Goodman R, Ford T, Simmons H, Gatward R, Meltzer H. Using the Strengths and Difficulties Questionnaire (SDQ) to screen for child psychiatric disorders in a community sample. Int Rev Psychiatry. 2003 Feb-May;15(1-2):166-72.
28. Meyer M, Elmer S, Jancke L. Musical expertise induces neuroplasticity of the planum temporale. Ann N Y Acad Sci. 2012 Apr;1252:116-23.
R e f e r ê n c i a s | 14
29. Hoenig K, Muller C, Herrnberger B, Sim EJ, Spitzer M, Ehret G, et al. Neuroplasticity of semantic representations for musical instruments in professional musicians. Neuroimage. 2011 Jun 1;56(3):1714-25.
30. Moreno S, Besson M. Influence of musical training on pitch processing: event-related brain potential studies of adults and children. Ann N Y Acad Sci. 2005 Dec;1060:93-7.
3 . A R T I G O S
A r t i g o s | 16
3.1 Music education for improving reading skills in children and adolescents with
dyslexia (Review)
A r t i g o s | 17
A r t i g o s | 18
A r t i g o s | 19
A r t i g o s | 20
A r t i g o s | 21
A r t i g o s | 22
A r t i g o s | 23
A r t i g o s | 24
A r t i g o s | 25
A r t i g o s | 26
A r t i g o s | 27
A r t i g o s | 28
A r t i g o s | 29
A r t i g o s | 30
A r t i g o s | 31
A r t i g o s | 32
A r t i g o s | 33
A r t i g o s | 34
A r t i g o s | 35
A r t i g o s | 36
A r t i g o s | 37
A r t i g o s | 38
A r t i g o s | 39
A r t i g o s | 40
A r t i g o s | 41
A r t i g o s | 42
3.2 EACOL (Scale of Evaluation of Reading Competency by the Teacher):
Evidence of Concurrent and Discriminant Validity
Published in the Journal of Neuropsychiatric Diseases and Treatment
EACOL (Scale of Evaluation of Reading Competency by the Teacher):
Evidence of Concurrent and Discriminant Validity
Hugo Cogo-Moreira1*, George B. Ploubidis3, Clara Regina Brandão de Ávila2, Jair de
Jesus Mari1, Angela Maria Vieira Pinheiro4
1. Department of Psychiatry, Federal University of São Paulo, São Paulo, Brazil
2. Department of Hearing and Speech Pathology, Federal University of São Paulo, São
Paulo, Brazil
3. Department of Population Studies, Faculty of Epidemiology and Population Health,
London School of Hygiene and Tropical Medicine, University of London, UK.
4. Department of Psychology, Federal University of Minas Gerais, Belo Horizonte,
Brazil.
*Corresponding Author: Hugo Cogo-Moreira, PhD Student, Department of Psychiatry,
Federal University of São Paulo. Rua Borges Lagoa, 570. 1°Andar, São Paulo, Brazil.
CEP: email: [email protected]; phone:+55 (11) 82083526
Email addresses of all authors:
Hugo Cogo-Moreira ([email protected])
George B. Ploubidis ([email protected])
Clara Regina Brandão de Ávila ([email protected])
Jair de Jesus Mari ([email protected])
Angela Maria Vieira Pinheiro ([email protected])
A r t i g o s | 43
Abstract
Aim: to provide information about the concurrent and discriminant validation of the Scale
of Evaluation of Reading Competency by the Teacher (EACOL), which is composed of
27 dichotomous items concerning reading aloud (17 items) and reading silently (10
items). Samples: three samples were used in this validation study. The first was
composed of 335 students with an average age of 9.75 years (SD = 1.2) from Belo
Horizonte (Minas Gerais State), where the full spectrum of reading ability was
assessed. The second two samples were from Sao Paulo city (Sao Paulo State), where
only children with reading difficulties were recruited. The first Sao Paulo sample was
labelled “SP-screening” (n = 617) with a mean age of 9.8 years (SD = 1.0), and the
other sample was labelled “SP-trial” (n = 235) with a mean age of 9.15 years (SD =
0.05). Method: results from a Latent Classes Analysis (LCA), in which two latent groups
were obtained as solutions, were correlated with direct reading measures. Also,
students’ scores on the Intelligence Quotient (IQ) test and on the Strengths and
Difficulties Questionnaire (SDQ) tested the discriminant validation. Results: latent
groups of readers underlying EACOL predicted all direct reading measures, while the
same latent groups showed no association with behaviour and intelligence
assessments, giving concurrent and discriminant validity to EACOL, respectively.
Conclusion: EACOL is a reliable screening tool which can be used by a wide range of
professionals for assessing reading skills.
Keywords: school children; latent class analysis; assessment; reading difficulties;
validation.
A r t i g o s | 44
Background
Evaluations and assessments of teachers are used to make educational
decisions regarding students and to provide feedback to them, as well as to parents and
school psychologists.1,2 Teachers’ reports can thus serve as a primary source of
information in the educational setting3 and play a very important role in assessment of
emergent literacy.4
The key issue that emerges in the educational context concerns the validity and
reliability of teachers’ evaluations and the contrast between this type of indirect
assessment, with direct forms involving the use of both behavioral methods and
structured tasks such as the number of correctly read words per minute from a list of
real words.4
A review of sixteen studies concerning the association between teachers’
evaluations and test scores obtained by students revealed a high level of validity for
teachers’ assessment measures, but at the same time, showed high variability in
reliability. The range of correlation for the indirect comparisons (teacher were asked to
use a rating of achievement in reading, math, social science, and language arts) was
0.28 to 0.86, whereas the direct tests (teachers were directly asked to estimate the
achievement test performance of their students; for example, the number of problems
on an achievement test that each student solved correctly) yielded a range from 0.48 to
0.92.3
Twenty years after this seminal work, a study showed that the predictive validity
of teachers’ reports for assessing emergent literacy skills of preschoolers was positive
with moderate to large effects, between teacher’s evaluation and children’s
performance.5 However, as for the tools with good psychometric evidences to assess
A r t i g o s | 45
children’s reading performance indirectly, there is a shortage of studies, in spite of the
increased demand for such instruments, especially those that can effectively identify
children at risk of future reading difficulties.6
Recently, in this direction, a study has showed that judgements of the teachers
about their students’s progress, based on a criterion-referenced assessment (children’s
phonic skills and knowledge), is better than most formal tests in the identification of
those who later experienced reading difficulties.7
An early identification of these problems through reliable measures with good
psychometric properties based on theoretical and empirical evidence may play a key
role in prevention. However, evidences about whether intervention could prevent the
development of dyslexia and/or reading comprehension impairment for children early
identified as at risk of reading difficulties it is limited.8
In Brazil, there is a lack of tools with good psychometric properties and with
theoretical foundations underlying the latent construct of reading using teachers’
evaluations. For example, a study found that teachers’ reports, although reliable as a
whole, failed to identify specific reading difficulties in a number of children, and
concluded that such conditions would only be detectable via functional analysis of the
reading processes9 or by offering teachers a criterion-referenced instrument to be taken
into account in their judment about the reading skills of students.10,11
EACOL
In order to implement such idea, the authors developed the Scale of Assessment
of Reading Competence of Students by the Teacher (in in Brazilian Portuguese: Escala
de Avaliação de Competência em Leitura de Alunos pelo Professor - EACOL) which
A r t i g o s | 46
evaluates reading aloud (RA) and silent reading (SR) of children in elementary school.
The preliminary version of EACOL was tested by De Salles and Parente in 2007 who
found significant associations between students’ performance in reading and writing
words (as well as text comprehension) and teachers’ perception of these skills via the
EACOL.12 The teacher, once assisted with a set of well defined criteria, becomes then
more capable of rating the reading and spelling performance of their students.
Development of the EACOL
To develop the EACOL, information such as the teacher’s experience, as well as
a literature review about word recognition and comprehension were taken into account.
Elements to formulate fifty-seven items which were thought to describe the reading
aloud (RA) and silent reading (SR) skills of elementary school children were obtained.11
An operational definition criterion to the classification of readers in three categories was
proposed: children who are 1) good, 2) not so good reader, and 3) poor readers. In
each category, the items were subdivided into items about reading aloud and silent
reading.
Ten independent experts (linguists and psychologists), specialists in
psychological assessment and development of reading were asked to evaluate the
relevance and applicability of items and criteria of reader ability. If the items were
pertinent, the experts were asked to give a grade from 1 to 5 (where 1 meant low and 5
very important relevance) to determine the importance level of each criteria in defining
evaluation of reading competency by teachers. In addition, the experts were invited to
suggest other relevant items and/or modifications to the list.
A r t i g o s | 47
Following this procedure, the EACOL had two forms: A and B. Form A was
developed to evaluate children’s reading skills in the final phase of literacy
(approximately seven years old). Form B was developed for children who are
completely literate. Recently, the Scale underwent to a final adjustment.13 In this paper,
we consider only form B (aimed at children from eight to ten years old).
Form B includes twenty-seven items, seventeen of which tap into the
competency of reading aloud. Six of these items are related to good reading ability, five
describe not so good reading competency, and six describe poor reading skills. The
remaining ten items focus on silent reading: four items tap good reading ability, three
items for not so good reading ability, and the three last items tap into poor reading
ability. The EACOL is in the appendix.
The EACOL’s has not yet been applied in any language other than the Brazilian
Portuguese. The translation into English was done and carried out into three steps by
the last author: translation from Portuguese to English, back translation by a linguist,
and correction and semantic adaptation when necessary.
Since the EACOL has not yet been submitted to a rigorous test, involving a large
sample of students, that tests both concurrent and discriminat validity (the former is
appropriate for test scores that will be employed in determining children’s current status
with regard to reading skills; the latter is the evidence based on the consistently low
correlation between measures that are supposed to differ [i.e., EACOL not being
correlated to psychiatric symptoms or intelligence]),14 the objectives of this study are:
(1) to identify subtypes of readers by means of the utilization of the EACOL; (2) to
describe the associations between the subtypes of readers found in this indirect
measure of reading and various measures directly related and unrelated to reading; and
A r t i g o s | 48
(3) to verify whether the EACOL is sensitive to changes in instructions.Therefore, our
hypotheses are: (1) If the EACOL is a useful screening tool for assessing reading skills
of 2nd–4th grade readers, latent groups of readers showing different levels of reading
ability will be found; (2) if the students judged by the teachers using the EACOL as
good, not so good, or poor readers show corresponding performance in direct measures
of reading, this will be taken as concurrent validity for the instrument. In the same
manner, discriminat validity will granted to it if no associations are found between the
reading ability of the sample and behavior and intelligence assessments; and (3) if the
EACOL is sensitive to changes in instructions, the number of latent groups found will
vary in accordance to the instructions given. That is, it is expected that a best-fit model
with three latent groups of readers will be obtained when no specific direction in
instruction is given to the teachers and that lesser latent groups will emerge in a
situation in which the teacher is explicitly asked to think of a particular type of reader,
whether good readers or poor readers.
Method
Sample Recruitment
Three samples were used in this validation study: one from Belo Horizonte
(Minas Gerais State) and two from Sao Paulo city (Sao Paulo State).
The first sample, which is called here BH-sample, is the main reference sample.
It was constituted by 335 children, students on average 9.75 years old (SD=1.2) from
second to fifth grades at five schools. Their teachers (N=23), who agreed to participate
through the Informed Consent, were asked to complete the EACOL under the following
instructions: “Could you please classify each of your students, according to the criterion
A r t i g o s | 49
presented? For each item please answer “Yes” if it describes the reading ability of the
student being evaluated and “No”, otherwise. Thank you for your collaboration.”
In the Sao Paulo’s samples, the main one was constituted by 235 children from
8- to 10-years old from ten public schools located in impoverished areas in the outskirts
of the city of Sao Paulo, which is part of a screening sample obtained from 617 children
(mean=9.8 years old (SD=1.0). 48 Teachers from the second to fourth grade of these
ten schools were asked to fill in the EACOL, considering only the children with “a
reading (ability) below the mean for the corresponding grade.” This instruction was
given to screen eligible children to take part in a randomized clinical trial about the
effectiveness of music education in the improvement of reading skills among children
with reading difficulties (clinicaltrial.gov: NCT01388881 and Research Ethic Committee
from Federal University of Sao Paulo CEP 0433/10). 617 children formed what we call
the Sao Paulo Screening Sample (SP-screening).
On the basis of the SP-screening, trained psychologists then ranked children who
had the worst scores in EACOL to identify, per school, a minimum of 24 children with
reading difficulties to participate in the randomized clinical trial about the effectiveness
of music education. Since the ten schools had different numbers of enrolled children,
four schools did not meet the criteria of 24 children per school. In the other six schools,
where the numbers of eligible children exceeded 24, a minimum of 24 and a maximum
of 27 children were randomly selected via a lottery. After having identified the eligible
children, the research team contacted the parents through a presentation letter with a
description of the objectives of the trial and the informed consent. In the case of interest
and acceptance by the parents, the children were considered included to participate.
A r t i g o s | 50
To avoid bias related to cognitive problems in the SP-trial due to the nature of the
experimental randomized clinical trial, the included children were tested for nonverbal
intellectual ability using the Raven’s Coloured Progressive Matrices;15,16 children with
scores below the 25th percentile were excluded. To avoid confounders due to
contamination or overlap of interventions, parents were asked if their child was already
receiving any regular hearing or speech therapy and/or music classes (such as private
music classes, a social project involving musical learning, or other music schooling).
The total number of eligible children who were indicated by the teacher, selected
by the psychologists as having the worst reading scores, and that returned the parents’
authorization was 240. Out of these, two children were excluded because they had a
score below the 25th percentile and three because they were already participating in
social projects which involved musical learning or/and were under regular consultation
with hearing and speech therapists. This left a sample of 235 children obtained from the
SP-screening (38.08%), who were classified as not so good and poor readers, with
average age 9.15 years (SD=.05). We call this group the Sao Paulo Trial Sample (SP-
trial).
Both the BH-sample and the SP-screening were taken as reference groups. Only
the SP-trial sample was submitted to the procedure described below – the study of the
external validation of the EACOL.
Measures
To evaluate the EACOL’s discriminant validity, we used Intelligence Quotient
total score (IQ)17 and the Strengths and Difficulties Questionnaire (SDQ) filled by
teachers,18 which is a brief behavioral screening questionnaire constituted by 25 items
A r t i g o s | 51
divided between 5 scales: emotional symptoms, conduct problems,
hyperactivity/inattention, peer relation problems and pro social behavior.
To test the concurrent validity of the EACOL, we selected a number of key
variables to act as outcomes in the reading domain. These included:
Accuracy in the word task (rate of correct real words read per minute) ;
Accuracy in the non-word task (rate of correct non-words read per minute) and.
Accuracy in the text task (rate of correct words read per minute);
The word and non-word tasks19 are consisted of a total of eighty-eight words and
eighty-eight non-words. The words varied in frequency levels of occurrence (high and
low frequency words),19 bi-directional regularity (regular and irregular words according
to grapheme-phoneme/phoneme-grapheme correspondence),20 and in length (short,
medium and long words, in terms of number of letters). The non-words were built with
the same Brazilian Portuguese orthographic structure and the same length of stimuli
used in the word list. Here only the total number of correct words and non-words read
per minute of these tasks will be considered; subgroup analysis related to regularity or
irregularity or even length’s word were not conducted.
Psychometrically, the word and non-words tasks showed excellent indices,
presenting high correlation between each other (r=0.92, p<0.001) and moderately
positive correlation with Phonological Awareness Test21 (r accuracy of word=0.40 and r accuracy
of non-word =0.37). As expected, the general Intelligence Quotient (IQ) was poorly related
to word accuracy (r= 0.168, p=0.01) and not correlated with non-word accuracy (r=0.01,
p=0.131). Also, via tobit regressions, adjusted for the clusters of 10 schools, schooling
effects through the academic years were observable in word accuracy (i.e., the higher
education level, the better was achievement in the accuracy of words (third grade
A r t i g o s | 52
β=6.62, p<0.01; and fourth grade β=10.56, p<0.01)) and accuracy of non-words (third
grade β=4.45, p<0.001; and fourth grade β=6.77, p<0.001), corroborating for internal
validation of both tasks.
Regarding the text that had to be read, a specific text was selected, considering
the age of the child. The accuracy of text reading correlated highly with word accuracy
(r=0.916; p<0.001) and with non-word accuracy (r=0.873; p<0.001).The children’s
reading was audio recorded for post-test analysis of accuracy.
Last, we included two covariates in the regressions models (described below).
The first was visual acuity (best-corrected, age-appropriate), under conditions of
monocular viewing, conducted by a technician in ophthalmology via Snellen´s Scale.
The children were classified as having visual alterations or not. The second was the
Simplified Central Processing Auditory Assessment22 conducted by a hearing and
speech pathologist; the following auditory abilities were tested: sound localization in five
directions; verbal and non-verbal sequential memory, corresponding to the processes of
localization and temporal time ordering. The children were classified as having
problems in central auditory processing or not.
Statistical Modeling
To verify the number of latent groups in the three samples (BH-sample, SP-
screening and SP-trial), we used Latent Classes Analysis (LCA) on the twenty-seven
dichotomous items of EACOL. The LCA is a form of cluster analysis initially introduced
by Lazarsfeld and Henry in 1968. It is the most commonly applied latent structure model
for categorical data23 allowing the specification of statistical distributions through a
model-based method which differs from methods that apply arbitrary distance metrics to
A r t i g o s | 53
group individuals based on their similarity (for example, K-means clustering).24 In the
LCA, unlike K-means clustering, a statistical model is built for the population from which
the data sample was obtained.25
To compare LCA models with different numbers of latent classes, we used the
Bayesian information criterion (BIC), in which small values correspond to better fit, as
well as the sample size-adjusted BIC (ssaBIC). The classification quality of the model
was evaluated with the entropy criterion, in which the values range from 0 to 1, where
values close to 1 indicate good classification. All LCA were conducted via Mplus version
6.12.26
Both samples from SP were used to test whether the type of instructions that
were given to the teachers would change the number of latent classes in comparison to
the BH reference sample. Both concurrent and discriminant validity were assessed
using a regressions model STATA version 12, considering robust standard errors to
adjust for the cluster structure. Covariates such as age, gender, grade, and visual acuity
and central auditory processing were considered in the regression models. It is
important to emphasize that distributions, kurtosis and skewness of the outcomes and
covariates were checked to choose the better regression model. To optimize the
visualization of the estimated probabilities as results of LCA, all “positive” EACOL items
(e.g., reads with intonation compatible with punctuation marks; quickly reads "new" and
invented words; quickly reads "known" words and "little known" ones) were reversed
scored for ease of interpretation. The recoded items were 3, 7, 12, 14, 15, 17, 20, 22,
23, and 25.
A r t i g o s | 54
Results
Results of the LCA
The LCA of BH suggested a good fit-model with three-classes, while two-class
model for SP-screening and SP-trial was confirmed.
Figure 1 – Latent Classes for BH-sample
A r t i g o s | 55
Figure 2 – Latent Classes for SP-screening
A r t i g o s | 56
Figure 3 – Latent Classes for SP-trial
To establish which class corresponds to which category of reader, it is necessary
to refer to the graph where the estimated probability axis has a scale from 0 to 1. The
former indicates good reading ability, while the later, represents reading disability.
In the BH-sample, a clear three-class solution is supported, considering empirical
and theoretical elements.
For SP-screening and SP-trial, the parametric bootstrap p-values for the
likelihood ratio X² goodness of fit test returned values of p<0.001 for the one, two, three,
and four-class models, but only the two-class solutions had theoretical and empirically
plausibility. Taking AIC, BIC, ssaBIC, and entropy results together with the theoretical
information about the EACOL and the principle of parsimony, the two-latent-class model
was deemed as the most appropriate to describe the data. The model identifies children
A r t i g o s | 57
groups with different patterns of reported reading in both these samples from Sao
Paulo.
Table 1 – Latent Classes Analysis Results
Abbreviations: AIC: Akaike Information Criterion; BIC: Bayesian Information Criterion; ssaBIC: sample size adjusted
Description of Typological (Latent) Classification
BH-sample
In this reference sample, a three class model (Figure 1) provided the optimal
solution, as observed in Figure 1. Class 1 had model-based prevalence of 26.9% of the
sample, class 2 had 12.1% and class 3 was the most prevalent at 61.1%. Class 2 is
represented by superior marginal probabilities (close to 1), class 3 inferior marginal
probabilities (close to 0) and class 1 is represented by the medial line where occur non-
BIC (L²) AIC (L²) ssaBIC Entropy
Parametric Bootstrapped
likelihood h0
Loglikelihood Value
p-value
Overall Bivariate
Log-likelihood
Chi-Square
BH-Sample
Model 1 8634.059 8531.158 8548.413 ---- ---- ---- 16102.95
Model 2 6562.997 6353.384 6388.533 0.987 -4238.579 <0.001 2868.681
Model 3 6058.233 5768.908 5821.95 0.975 -3121.692 <0.001 770.341
Model 4 6026.771 5603.735 5674.67 0.982 -2801.454 <0.00001 673.126
SP-Screening
Model 1 18317.67 18198.19 18231.95 ---- ---- ---- 5558.043
Model 2 17173.98 16930.62 16999.37 0.819 -9072.097 <0.001 1198.651
Model 3 17089.5 16722.24 16825.99 0.854 -8410.308 <0.001 832.6
Model 4 17078.06 16586.9 16725.66 0.811 -8278.119 <0.00001 686.545
SP-trial
Model 1 6336.019 6242.61 6250.44 ---- ---- ---- 1365.364
Model 2 6146.722 5956.495 5972.455 0.827 -3094.305 <0.001 589.026
Model 3 6200.666 5913.52 5937.591 0.853 -2923.247 <0.001 456.106
Model 4 6253.835 5869.821 5902.012 0.889 -2874.432 <0.001 395.367
A r t i g o s | 58
marginal probabilities (the majority of probabilities are centred between 0.25 and 0.75).
There are three distinct lines which have a small amount of overlap and only two
crossed-trajectories (items 16 and 17). In the BH, we refereed to class 3 as good
readers class 1 as not so good readers and class 2 as poor readers.
Sao Paulo’s samples
The latent structure of the classes in both samples was similar, considering the
distribution of estimated probabilities through the twenty-seven items, the number of
classes, and the proportion between the percentages of children in each class.
In the graph of the estimated probabilities for SP-screening sample (Figure 2),
class 1 comprised 39.7% of the sample. In the case of the SP-trial (Figure 3), class 1
had a model-based prevalence estimate of 37.2% and included children with median
probabilities (from 0.3 to 0.7) in the majority of items of reporting. We referred to this
class as “not so good readers.”
Class 2 in the SP-trial had a model-based prevalence of 62.8% and included
children who had a marginal value (p>0.8) probability, which indicates that eleven out of
the twenty-seven EACOL items were applicable to describe poor reader´s class. The
highest probabilities were observed in the following items: “can summarize the text read
orally (item 20)”, “can identify characters, places and ideas in the main text, after the
first reading (item 25 )” and “quickly reads "known" words and "little known" ones
(item12)”. In the SP-screening sample, the percentage of the children in class 2 was
60.3%.
The two samples from Sao Paulo returned similar prevalence for the two latent
classes. In addition, it is possible to observe that some items have results with “crossed
A r t i g o s | 59
trends” or even “overlapped trends.” This means that these items are not good for
discriminating classes, and therefore in later studies, they could be omitted or excluded.
Regarding to both reading domains (AR and SR) separately, while in BH-sample
RA (from item 1 to item 17) works better, in the samples from Sao Paulo SR domain
(from item 18 to item 27) distinguishes better the poor reader from the not so good
reader class.
Discriminant and Concurrent validity
IQ and SDQ
To test EACOL’S discriminant validity we used the general IQ and total difficulties
children’s score in the SDQ as a dependent variable against the same exploratory
variable was used in the above-cited model. We did not observe an association
between the latent classes and IQ (β=6.55, p>.05) and total difficulties children’s score
measured by the SDQ (β=1.12, p>.05), as observed in the Table 2, when controlled by
age, gender, grade and school as cluster unity.
Reading Outcomes
In the regression analyses, the class latent typology had a significant negative
association with the three reading measures, controlling for age, grade, gender, and
visual acuity and processing auditory status; also the cluster design was considered and
as consequence robust standard errors were generated. Results are described in the
Table 2.
A r t i g o s | 60
Being a member of poor reader class has major negative impact in all reading
outcomes, showing that this group has more reading difficulties than not so good
reader. For the accuracy of words (β=-11.12, p<.0001), (in other words, a significant
difference of 11 correctly read words per minute between both latent groups of readers)
and accuracy of non-words (β=-6.50 p<.001), we used tobit regression due to the floor
effects in both continuous outcomes (children who have read zero words\non-words
correctly) and, therefore, we specified one left-censoring limit of 1correct read word per
minute. For the accuracy of text reading (β=-11.27, p<.01) a linear regression model
was used, showing that there is an effect in being class 1 or 2 on the outcome. More
precisely, comparing not so good reader and poor reader, we expected that the worst
indicators of reading were achieved from poor readers (Class 2).
Table 2 – Values for Regression Coefficients with its respective robust standard error, p-
value and 95% confidence interval for variables of Concurrent and Discriminant Validity
Outcomes on two Latent
Groups from LCA Coef. (β)
Robust Std. Err.
t P-value [95% Conf.
Interval]
Concurrent Validity
Accuracy of non-word -6.50 1.546 -4.21 <0.001 -9.55 -3.45
Accuracy of Word -11.12 2.672 -4.16 <0.001 -16.39 -5.85
Accurracy of Text -11.27 3.732 -3.02 0.014 -19.71 -2.83
Discrimant Validity
IQ Total 1.12 0.845 1.33 0.217 -0.79 3.03
Total Difficulties Score (SDQ) -1.60 1.604 -1.00 0.344 -5.23 2.03
Discussion
The present study explored the predictive ability of indirect measures of teachers’
reports on children’s reading ability. The latent groups of readers predicted direct
measures of reading abilities, particularly in the area of decoding of isolated words
(represented here as accuracy of word and non-word reading) and words in context
A r t i g o s | 61
(represented as the accuracy of reading text). Considering the SP-trial, the poor reader
latent group correctly read 6.50 non-words less per minute than the not so good reader.
Also, in the other reading measures, the differences between both groups are
statistically significant: The poor reader latent group’s performance was worse than that
of the not so good reader group regarding the accuracy of reading both isolated words
and words in context (difference of 11.12 and 11.27 correctly read words per minute,
respectively) (Table 1). These results are evidence of concurrent validity of the EACOL.
We also evaluated the extent to which the instructions given to the teachers
could accurately affect the identification of the latent groups of readers.
Major Findings and Clinical Implications
The BH sample returned a three-class model while SP samples returned a two-
class model due to the instructions that were given to the teachers. As a consequence,
the number of returned classes must be different, giving evidence for concurrent validity
of EACOL. Considering that EACOL in the BH-sample covered the full spectrum of the
reading abilities (i.e., no discriminative instruction was given to the teachers), BH-
sample prevalence results may suggest that either the teacher has a tendency to
overestimate the children’s reading ability, or perhaps teachers tend to or even prefer to
answer about children who have non-specific academic difficulties of some sort. Since
EACOL inquires about specific characteristics of children’s reading, which normally are
observable one-by-one, it is necessary to have a proximal contact with the child,
especially to evaluate items related to silent reading, which showed better discrimination
in SP-sample (i.e., samples of children with reading difficulty). Therefore, when no
specification of the type of reading ability of the participants is requested, teachers may
A r t i g o s | 62
tend to complete the EACOL considering predominantly the children with good and not
so good reading abilities, which were the major prevalence groups in the BH-sample.
We expected to observe proportions among the three classes to be similar to the
normal curve, where the majority of the children would be categorized around the mean
(corresponding to the average reader, here the “not-so-good reader”) and the minority
(both good and poor readers) would be placed with regard to the marginal probabilities,
closer to 0 and closer to 1, respectively.
The BH sample returned a three-class model, while SP samples returned a two-
class model due to the instructions given to the teachers. As a consequence, the
number of returned classes must be different, giving evidence for concurrent validity of
EACOL.
As it is possible to see in Table 1, the entropy value (i.e., how well the classes
are distinguished from each other) in both samples from Sao Paulo are lower than BH.
This could refer to the difficulty of teachers in evaluating children due to the instructions,
especially those who had reading difficulties. Taking into consideration the BH-sample,
a three-class solution was achieved and only two items had “crossed values.”
Therefore, distinctions between the three categories when the teachers are free to
consider the full spectrum of reading (and readers) make the distinction more precise.
Taking the two domains of the scale RA and SR into account, some details could
be addressed about the twenty-seven items which are divided in the LCA graphs. In the
case of SP-trial and SP-screening, it is possible to observe in the graphs of estimated
probabilities a major overlapping in the RA domain (i.e., represented by two lines closer
or with the same trajectory), while in the SR the two lines do not overlap.
A r t i g o s | 63
In the BH-sample, RA works better than SR in which the good reader and not so
good reader (respectively classes 3 and 1 in the Figure 1) have very close probabilities
(p<0.1), taking into account five of the ten items (items 19, 20, 21, 22, 23). In the RA,
the item 15 (reads with rhythm, not too slowly nor too quickly) did not have the
probability to discriminate the classes of good reader and not so good reader, because
there was an overlap between two classes in the same probability. With the exception of
this item, RA seems better differentiate the classes when no direction is given to teachers,
probably because in the school context it might be easier to observe difficulties in
reading aloud than in silent reading, as evaluation of this later type of reading is often
obtained “head-to-head”, through specific investigation and inquiry about the students’
comprehension capacity (such as his/hers ability to use knowledge of world, to make
inferences and monitor the understanding of what is being read).27 On the other hand,
when teachers were required to think about the children with reading difficulties, silent
reading become a better measure to distinguish not so good readers from poor readers
since that in this condition, the overlapping of trajectories among the items are less
frequent. With respect to discriminant validity, the latent groups in SP-trial did not
predict either in IQ, as found also by Hatcher & Hulme,28 or SDQ as expected, showing
that both domains were not associated with reading skills. More specifically, children’s
behavioural characteristics evaluated via SDQ were not taken into account in teachers’
evaluation of children’s reading. In other words, teachers were capable of distinguishing
presence of behavioural problems from reading difficulties, indicating that both
theoretical constructs domains were independently evaluated by them. This is in
disagreement with the finding that the teachers’ perceptions of their students' behavior
constituted a significant component of the judgments made about their students
scholastic achievements.29
A r t i g o s | 64
Therefore, this study found evidence that the EACOL is a reliable instrument to
assess reading via the teachers’ judgment. Since it is simple and easy to administer, it
is an important tool to help a wide range of professionals (e.g., health professionals who
work with children, teachers and educators, as well as researchers).
Acknowledgement
This study was funded by the National Council for Scientific and Technological
Development (CNPq – grant n° 482321/2010-5) and the Instituto ABCD, which is a non-
governmental organization (NGO) that supports research about dyslexia in Brazil.
Ethical Issues
This protocol for the randomized clinical trial (SP-screening and SP-trial) was
submitted to and approved by the Ethical Research Committee of the Federal University
(CEP0433/10) of São Paulo (UNIFESP). The protocol for BH sample was approved by
the Ethical Committee from the Federal University of Minas Gerais (Process: n ETIC
347/04).
A r t i g o s | 65
Appendix EACOL Form B
Student’s name: _________________________________school year: _____
Age: ____years ____months Teacher’s name:_______________________
School’s name ___________________________________________________
Evaluation of reading aloud
Nº Subtypes of
readers
Criterion Yes No
01 Poor reader Reads but cannot tell what was read, even when stimulated
with questions.
02 Not so good
reader
Sometimes makes mistakes when reading "new" words.
03 Good reader Reads with intonation compatible with the punctuation marks,
expressing emotions and feelings according to the text read.
For example, gives an intonation of questioning in the whole
sentence, when there is question mark in the text. Give
intonation of joy or surprise, in the whole sentence, when
there is an exclamation mark.
04 Poor reader Does not take into account the intonation compatible with the
punctuation marks, reading in a monotone manner.
05 Poor reader Says "I do not know" when encounters a new word.
06 Not so good
reader
Sometimes reads and cannot tell what was read.
07 Good reader Quickly reads "new" and invented words.
08 Poor reader Reads very slowly, without rhythm, spelling out each syllable,
does not observe the punctuation marks.
09 Poor reader Reads spelling out both "new" and "known" words.
10 Not so good
reader
Sets the tone of interrogation and / or exclamation only in the
word that precedes the punctuation mark.
11 No so good
reader
Delay start reading when "new" words are encountered,
needing to spell them out.
12 Good reader Quickly reads the "known" words and the "little known" ones.
13 Poor reader Often makes mistakes when reading "new" words.
A r t i g o s | 66
14 Good reader Seems to have understood what was read when asked about
the text read.
15 Good reader Reads with rhythm, not too slowly nor too fast.
16 Not so good
reader
Reads too slowly or too quickly.
17 Good reader Reads words correctly.
Evaluation of Silent Reading
Nº Criterion Yes No
18 Not so good
reader
Does identify characters and places, but has some difficulty to
identify main ideas without a second reading.
19 Poor reader Does not identify the subject from the title, nor vice versa.
20 Good reader Can summarize the text read orally.
21 Poor reader Does not identify characters, places or main ideas expressed
in the text.
22 Good reader
Is able to choose a title for passages presented with no title or
even an alternate title for titled passages.
23 Good reader Is able to identify the subject from the title and vice versa.
24 Not so good
reader
Presents some difficulty in orally summarizing what was read.
25 Good reader Can identify characters, places and ideas of the main text, after
the first reading.
26 Not so good
reader
Not always able to identify the subject from the title and vice
versa.
27 Poor reader Not able to summarize what was read, either orally or in
writing.
A r t i g o s | 67
References
1. Elliott SN, Gresham FM, Freeman T, McCloskey G. Teacher and observer ratings of
children's social skills: Validation of the Social Skills Rating Scales. J Psychoeduc
Assess. 1988;6(2):152-161.
2. Gerber MM, Semmel MI. Teacher as imperfect test: Reconceptualizing the referral
process. Educ Psychol. 1984;19(3):137-148.
3. Hoge RD, Coladarci T. Teacher-Based Judgments of Academic-Achievement - a
Review of Literature. Rev Educ Res. Fal 1989;59(3):297-313.
4. Salinger T. Assessing the literacy of young children: The case for multiple forms of
evidence. In: Neuman SB, Dickinson DK, eds. Handbook of early literacy research.
Vol 1. New York: Guilford Press 2002:390-418.
5. Cabell SQ, Justice LM, Zucker TA, Kilday CR. Validity of Teacher Report for
Assessing the Emergent Literacy Skills of At-Risk Preschoolers. Lang Speech Hear
Ser. Apr 2009;40(2):161-173.
6. Schatschneider C, Petscher Y, Williams KM. How to evaluate a screening process:
The vocabulary of screening and what educators need to know. In: Justice LM,
Vukelich C, , eds. Achieving excellence in preschool literacy instruction New York:
Guilford Press; 2008:304–316.
7. Snowling MJ, Duff F, Petrou A, Schiffeldrin J, Bailey AM. Identification of children at
risk of dyslexia: the validity of teacher judgements using 'Phonic Phases'. J Res
Read. May 2011;34(2):157-170.
8. Snowling MJ, Hulme C. Annual Research Review: The nature and classification of
reading disorders - a commentary on proposals for DSM-5. J Child Psychol Psyc.
May 2012;53(5):593-607.
9. Pinheiro AMV. Heterogeneidade entre Leitores Julgados Competentes pelas
Professoras. Psicol. Refl. Crít. 2001;14(3):537-551.
A r t i g o s | 68
10. Pinheiro AMV. Os passos da construção da Escala de Avaliação de Competência
de Leitura de alunos pelo professor – EACOL Paper presented at: VII Internacional
de Investigação em Leitura, Literatura Infantil e Ilustração Meeting2012; Braga,
Lisbon.
11. Pinheiro AMV, Costa A. Escala de avaliação de competência em leitura pelo
professor. Paper presented at: VII Encontro Mineiro de Avaliação Psicológica2005;
Belo Horizonte, Brazil.
12. De Salles JF, Parente MAMP. Relação entre desempenho infantil em linguagem
escrita e percepção do professor. Cad Pesqui. 2007;37(132):687-709.
13. Pinheiro AMV. Validação e estabelecimento de normas de uma prova
computadorizada de reconhecimento de palavras para crianças. Belo Horizonte:
Federal University of Minas Gerais;2012. SHA - APQ-01914-09.
14. Urbina S. Essentials of Psychological Testing. Hoboken, New Jersey: John Wiley &
Sons; 2004.
15. Pasquali L, Wechsler S, Bensusan E. Matrizes Progressivas do Raven Infantil: um
estudo de validação para o Brasil. Aval. psicol. 2002(2):95-110.
16. Raven J. Raven’s progressive matrices and vocabulary scales. London: Lewis;
1986.
17. Wechsler D. WISC-III: Wechsler intelligence scale for children. San Antonio: The
Psychological Corporation.; 1991.
18. Goodman R. The Strengths and Difficulties Questionnaire: a research note. J Child
Psychol Psychiatry. Jul 1997;38(5):581-586.
19. Pinheiro AMV. Contagem de frequência de ocorrência de palavras expostas a
crianças na faixa pré-escolar e séries iniciais do 1º grau. Sao Paulo: ABD-
Associação Brasileira de Dislexia; 1996.
A r t i g o s | 69
20. Pinheiro AMV. Anexo 2. In: Sim-Sim I, Vianna FL, eds. Para a avaliação do
desempenho de leitura. Lisbon: Gabinete de Estatística e Planeamento da
Educação; 2007:121-131.
21. Capovilla AGS, Capovilla FC. Prova de consciência fonológica: desenvolvimento de
dez habilidades da pré-escola à segunda série. Temas desenvolv. 1998;7(37):14-
20.
22. Engelmann L, Ferreira MIDC. Avaliação do processamento auditivo em crianças
com dificuldades de aprendizagem; Auditory processing evaluation in children with
learning difficulties. Rev. soc. bras. fonoaudiol. 2009;14(1):69-74.
23. Clogg CC. Latent class models. New York: Plenum; 1995.
24. Everitt BS, Hand DJ. Finite mixture distributions. Monographs on Applied Probability
and Statistics, London: Chapman and Hall. 1981;1.
25. Magidson J, Vermunt J. Latent class models for clustering: A comparison with K-
means. Canadian Journal of Marketing Research. 2002;20(1):36-43.
26. Muthén L, Muthén B. Statistical analysis with latet variables: use's guide. Fourth ed.
Los Angeles, CA 1998-2011.
27. Ricketts J. Research Review: Reading comprehension in developmental disorders
of language and communication. Journal of child psychology and psychiatry, and
allied disciplines. Nov 2011;52(11):1111-1123.
28. Hatcher PJ, Hulme C. Phonemes, Rhymes, and Intelligence as Predictors of
Children's Responsiveness to Remedial Reading Instruction: Evidence from a
Longitudinal Intervention Study. J Exp Child Psychol. 1999;72(2):130-153.
29. Bennett RE, Gottesman RL, Rock DA, Cerullo F. Influence of behavior perceptions
and gender on teachers' judgments of students' academic skill. J Educ Psychol.
1993;85(2):347.
A r t i g o s | 70
3.3 Effectiveness of Music Education for the Improvement of Reading Skills and
Academic Achievement in Young Poor Readers: a Pragmatic Cluster-
Randomized, Controlled Clinical Trial
Aceito no Plos One (18 de fevereiro 2013)
Hugo Cogo-Moreira 1*, Clara Regina Brandão de Ávila 2, George B. Ploubidis 3, Jair de
Jesus Mari 1
1. Department of Psychiatry, Federal University of Sao Paulo, Sao Paulo, Brazil
2. Department of Hearing and Speech Pathology, Federal University of Sao Paulo, Sao
Paulo, Brazil
3. Department of Population Studies, Faculty of Epidemiology and Population Health,
London School of Hygiene and Tropical Medicine, University of London, UK.
*Corresponding Author: Hugo Cogo-Moreira, PhD Student, Department of Psychiatry,
Federal University of São Paulo. Rua Borges Lagoa, 570. 1°Andar, São Paulo, Brazil.
CEP: email: [email protected]; phone:+55 (11) 82083526
Email addresses of all authors:
Hugo Cogo-Moreira ([email protected])
Clara Regina Brandão de Ávila ([email protected])
George B. Ploubidis ([email protected])
Jair de Jesus Mari ([email protected])
A r t i g o s | 71
Abstract
Introduction: difficulties in word-level reading skills are prevalent in Brazilian schools
and may deter children from gaining the knowledge obtained through reading and
academic achievement. Music education has emerged as a potential method to improve
reading skills because due to a common neurobiological substratum.
Objective: to evaluate the effectiveness of music education for the improvement of
reading skills and academic achievement among children (eight to 10 years of age) with
reading difficulties. Method: 235 children with reading difficulties in 10 schools
participated in a five-month, randomized clinical trial in cluster (RCT) in an impoverished
zone within the city of São Paulo to test the effects of music education intervention while
assessing reading skills and academic achievement during the school year. Five
schools were chosen randomly to incorporate music classes (n=114), and five served
as controls (n=121). Two different methods of analysis were used to evaluate the
effectiveness of the intervention: The standard method was intention-to-treat (ITT), and
the other was the Complier Average Causal Effect (CACE) estimation method, which
took compliance status into account.
Results: The ITT analyses were not very promising; only one marginal effect existed for
the rate of correct real words read per minute. Indeed, considering ITT, improvements
were observed in the secondary outcomes (slope of Portuguese=0.21 [p<0.001] and
slope of math=0.25[p<0.001]). As for CACE estimation (i.e., complier children versus
non-complier children), more promising effects were observed in terms of the rate of
correct words read per minute [β=13.98, p<0.001] and phonological awareness [β
=19.72, p<0.001] as well as secondary outcomes (academic achievement in
A r t i g o s | 72
Portuguese [β =0.77, p<0.0001] and math [β=0.49, p<0.001] throughout the school
year).
Conclusion: the results may be seen as promising, but they are not, in themselves,
enough for make music lessons as public policy.
Keywords: randomized clinical trial; effectiveness; music education; reading difficulties;
academic achievement; structural modeling equation
A r t i g o s | 73
Introduction
Due to the demands of an increasingly technological society, reading
failure has a major impact on cognitive development [1,2]. Obtaining adequate reading
comprehension of written material is the ultimate goal of reading, and achievement of
word-level skills is used as an initial indicator of success in learning to read [3]. In 2009,
Brazil was ranked 53rd among 65 participating countries in reading and science
achievement and 57th in math via the Programme for International Student Assessment
(PISA) by the Organization for Economic Co-Operation and Development. Though PISA
analyzed 15-year-old children (an older population when compared with our sample of
8- to 10-year -olds), these indicators warrant attention from authorities not only in Brazil
but also in other countries with low achievement (e.g., Peru, Panama, Montenegro,
Bulgaria, and the Russian Federation).The most common approach to reading
intervention has a theoretical motivation: Good phonological and metaphonological
skills are important for success in learning to read. Children who have reading
difficulties have deficits in these skills and training in phonological skills in the context of
reading has repeatedly been shown to lead to improvement in reading, at least in
English [4].
Musical learning has emerged as a possible intervention due to the similarities
between musical learning—a non-verbal language—and verbal language itself. In
particular, musical learning can assist in the processing of lexical skills [5] and in
improving pitch discrimination abilities in both speech and reading among non-musician
children [6]. Cross-sectional studies have shown that the detection of pitch patterns
(global structure) is predictive of performance on measures of phonological skills and
reading ability [7]. Meanwhile, the structural development of the auditory cortex is
influenced by early musical experience [8]. Additionally, it has been pointed out that a
A r t i g o s | 74
link exists between musical abilities and phonological skills [9]; however, the bases of
these links are not clear [10].
The explanation of the causal paths to reading development via musical training
may be referred to as “transfer” [11,12]. The connection between musical learning and
improving reading skills would be a “far transfer” because musical learning is not directly
related to reading. Musical training is based on teaching and constant practice of non-
verbal structures such as classical sheet music, while reading is verbal. An example of
a “near transfer” would be learning to play a musical instrument and consequently
developing motor skills.
Neuroimaging studies have shown that some cognitive functions, such as the
ability to organize isolated words into meaningful sentences and the ability to organize a
variety of musical notes into a melody, may involve common neural pathways for both
speech and music [13].
Music education classes involve different cognitive functions that require complex
auditory pattern-processing mechanisms, attention, memory storage and retrieval,
motor programming, and sensory–motor integration [14]. However, a recent systematic
review of the effectiveness of music education used terms including “dyslexia” and
“reading difficulties/disabilities” and returned 876 citations, from which no randomized
clinical trials (RCT) were found. Therefore, despite the fact that musical learning is
popular and considered to be a beneficial intervention, there is no evidence from
randomized controlled trials that demonstrates the potential advantages of music
education on reading skills and consequently on academic achievement [15].
This research used a pragmatic RCT to address the effectiveness of music
education for improving reading skills and academic achievement in children with
reading difficulties, aged eight to 10. The main idea behind this pragmatic RCT was to
A r t i g o s | 75
reflect the heterogeneity of children with reading difficulties in the general public
education system, minimizing, as a consequence, exclusion criteria and providing a
more realistic scenario due its good external validity (generalizability of the results) [16].
The study aimed to test the effectiveness of music education classes for
improvement of academic achievement (based on Portuguese and math grades) and
word-level reading skills among children with reading difficulties. This trial is registered
at ClinicalTrial.gov under the number NCT01388881.
Method
Recruitment
School selection - inclusion criteria
Two Brazilian non-governmental organizations, or NGOs (specifically,
Partnerships of Education and Rukha’s Institute) that worked in impoverished
neighborhoods in Sao Paulo city (e.g., in slums) assisted in selecting 10 public schools
on the outskirts of the city. These schools were chosen based on several logistical and
social factors:
At least, one room available for music lessons. This room would also be
needed for the team of psychologists, audiologists, and ophthalmologists
to evaluate the children during the screening process and outcome
assessments;
The schools lacked music lessons in the curriculum.
Children’s selection
Inclusion criteria
Teachers from the second to the fourth grades of these schools were asked to
complete the Scale of Assessment of Reading Competence by the Teacher (EACOL)
which contains 27 dichotomous items with good divergent and concurrent validity,
A r t i g o s | 76
evaluates the loud (17 items) and silent reading abilities (10 items) of elementary school
children [17]. EACOL has a range of 29 to -29 points, where values closer to 29
represent a good reader, and the following cut-off scores were used to separate
students into three categories: the poor reader (<-14.5), not-so-good reader (from 14.5
to -14.5), and good reader (>14.5).
The following instructions were given: “…for the children in your class with a
reading ability below the mean for the corresponding grade, please fill out the EACOL.”
A total of 733 EACOLs from 48 teachers were returned, but only 617 were considered
valid. EACOLs were omitted if items were filled out inadequately— for example, there
were more than two missing items or sequential answers in a single category, or
teachers answered “yes” to all 27 items or “no” to all 27 items. The 617 valid children
formed what we labeled the Sao Paulo Screening Sample (SP-Screening). On the basis
of the SP-screening, the psychologists ranked the children who were classified as poor
readers or not-so-good readers in order to identify a minimum of 24 and a maximum of
27 children with reading difficulties to participate in the (RCT) from each school.
Because the 10 schools differed in their numbers of enrolled children, four schools did
not meet the minimum criteria. In the other six schools, where the numbers of eligible
children exceeded 24, a total of 27 names were randomly selected via a lottery. We
allocated a maximum number of students in order to prevent likely dropouts during the
academic year or loss due to exclusion criterion, which is described below.
After identifying the eligible children via the EACOL, the research team contacted
the parents via a letter that described the objectives of the trial. The letter explained the
study' aims, procedures, measurements, avoiding technical scientific vocabulary;
together with it, it was requested the parents' written informed consent which was
approved by Ethical Committee from Federal São Paulo University (CEP0433/10) for
A r t i g o s | 77
their children's participation. The Ethical Committee from Federal Sao Paulo University
approved this consent procedure. Only the children whose parents gave the written
consent were included in the study. All written informed consents were stored in the
department of Psychiatry at São Paulo Federal University. This study was approved by
the Ethical Committee from Federal São Paulo University.
Exclusion Criterion
To avoid bias related to cognitive problems, the included children were
tested for non-verbal intellectual ability using the Raven’s Coloured
Progressive Matrices [18], and children with scores below the 25th
percentile were excluded;
To avoid confounders due to contamination or overlap of interventions,
parents were asked if their children already were receiving any regular
hearing or speech therapy and/or music classes (such as private music
classes, social projects involving musical learning, or other music school
experiences). Children participating in such programs were excluded from
the study.
Sample Size
In total, 240 children were eligible for the study after being chosen by their
teachers; selected by the psychologists as having the worst reading scores; and
authorized by their parents to participate in the study. This value was based on the
sample size calculation, with the following points taken into account:
a) the cluster two-level structure (i.e., children who are nested in the schools);
b) the necessary number of children in each of the 10 schools selected to
achieve the minimum statistical power (1-β) of 0.75;
c) two measures (pre- and post-test assessments of the primary outcome); and
A r t i g o s | 78
d) the following parameters: ρ (rho – expected intraclass correlation coefficient
(ICC))=0.025; the expected moderate effect size (δ=0.45); α=0.05; and J (number of
clusters)=10.
The number of children per school was 24, with 240 children in the total sample.
From these 240 children, three were excluded because their parents retracted consent
after the full assessment of primary and secondary outcomes, and two changed schools
before the full reading evaluation took place. The in-cluster structure is also in
accordance with pragmatic design, reflecting the reality of the educational system.
Ultimately, a sample of 235 children (girls=38.3%) with an average age of 9.15 years
(SD=.05) was obtained from the SP-screening. The description of the above cited
process can be found in the flow chart diagram.
Measures
Potential Confounders
Before the assessment of the primary and secondary outcomes occurred, the
following were assessed in order to avoid confounders:
The visual acuity of the children (age-appropriate) under conditions of monocular
viewing, conducted by an ophthalmology technician using Snellen’s chart. The
children were classified as either having visual alterations or not. Also, auditory
processing was evaluated via the Simplified Auditory Processing Test (SAPT)[19]
by a hearing and speech-language pathologist. The following auditory abilities
were tested: sound localization in five directions; verbal and non-verbal
sequential memory; and the elicitation of the auropalpebral reflex through
instrumental sounds. The children were classified as having or not having
problems in central auditory processing.
A r t i g o s | 79
The intelligence quotient (IQ) was measured by a trained psychologist using the
complete Wechsler Intelligence Scale for Children—Third Edition (WISC-III)
[20,21].
School background variables were collected, including the number of classmates
of each included child and the annual presence of children in official classes.
Primary Outcome
To measure children’s’ ability to analyze metaphonological skills, the Test of
Phonological Awareness [22] was utilised. It consists of 10 subtests, each one featuring
four items used to verify synthesis, segmentation, manipulation, syllabic transposition,
phonemic synthesis, rhyme, and alliteration. Therefore, the score range was from 0 to
40.
Phonological awareness strongly predicts reading skills [23] and is widely accepted
to be an area of deficit among dyslexic children [24,25]. Reading is a complex and
multivariate process, and so we focused on variables related to lower-level cognitive
skills (word-level reading) as our primary outcomes. The measured skills included the
following:
A word accuracy task (rate of correct real words read per minute),
A non-word accuracy task (rate of correct non-words read per minute) and
An in-text accuracy task (rate of correctly read words per minute in the text).
The lists were used for the first time in this trial and included 88 words and 88 non-
words. The words varied in occurrence frequency (high- and low-frequency words), bi-
directional regularity (regular and irregular words according to grapheme-
phoneme/phoneme-grapheme correspondence); and length (short, medium, and long
words, as measured by the number of letters). The non-words were built with the same
orthographic Brazilian Portuguese structure, and the same length of stimuli was used in
A r t i g o s | 80
the list of words. Psychometrically, the word and non-words tasks showed excellent
indices, presenting high correlations (r=0.92, p<0.001). In addition, both were correlated
positively and moderately with phonological awareness (r word accuracy =0.40 and r non-word
accuracy =0.37). As expected, the general Intelligence Quotient (IQ) was related poorly to
word accuracy (r= 0.168; p=0.01) and not correlated with non-word accuracy (r=0.01;
p=0.131).
Regarding the text-reading task, three different texts were selected for the three
different age groups. The baseline in-text accuracy correlated highly with word accuracy
(r=0.916; p<0.001) and with non-word accuracy (r=0.873; p<0.001).
In all of the above situations, the children’s reading was audio-recorded for
accuracy analyses. The researchers had intended to blind the speech-language
pathologists who collected the primary outcome data, but during the second evaluation,
comments about the study allocation from teachers, directors and from the own children
make the speech-language pathologists discover about the status of school as
intervention or control.
The secondary outcome was academic achievement based on Portuguese and
math grades. These were measured four times by the teachers during the school year,
which begins in February and ends in November. The school directors were contacted
at the end of school year to collect the Portuguese and math grades from the children in
the trial. The grades were measured from 0 to 10, with 10 being the highest possible
grade. None of the school directors or teachers were blinded to the randomization
status of the school.
The Randomization Procedure
In July 2011 (the middle of the school year in Brazil), the 10 directors of the 10
schools were invited to participate in a lottery. Two opaque boxes were used: The first
A r t i g o s | 81
contained balls containing ordinal numbers from one to 10. The numbers that the
directors picked corresponded with the sequence of the subsequent lottery. The second
box contained five balls printed with the word “intervention,” and five others were printed
with “control.” In a sequence determined by the ball number picked in the first lottery,
each director was called to pick one ball from the second box—either a “control” ball or
an “intervention” ball. For example, the director who picked the ball with the number five
in the first lottery was the fifth to pick either a control or intervention ball from the second
box. Because we worked with a purposeful sampling of the schools, the randomization
procedure was important for excluding bias related to school selection.
Intervention
Music education (briefly defined here as a process of musical learning) was
methodologically and educationally based on Brazil’s National Curriculum Parameters
(NCP) [26]. This program focuses on a modern approach to music education in which
the process of musical learning is not restricted to the domain of Western and classical
sheet music reading or to a high aptitude for a particular musical instrument. Rather, the
program focuses on musical improvisation, composition, and interpretation in
accordance with the National Association for Music Education [27].
Children were encouraged to create their own music and to perceive and identify
musical elements (rhythm, melody, harmony) during 50-minute activities that occurred
three times per week for five months starting at the end of June 2010 and ending the
last week of October 2010. Children were called to create and play music as well as to
explore the sounds and history of non-traditional classical instruments made for avant-
garde musical compositions and composers of the 20th century. Each school received
soprano and contralto block flutes, keyboards, and two music teachers.
A r t i g o s | 82
All music teachers followed the same syllabus and musical activities to avoid
educational bias and to make the classes as similar as possible. The teachers were
randomly allocated to the five intervention schools. Every two weeks during the
intervention period, supervisions were arranged with the researchers, who
systematically verified whether the music teachers were following the NCP’s
assumptions and educational structure. Two teachers were provided per class to
improve children’s level of attention and to guarantee that if any music teacher was
absent, the other would follow the pedagogical plan. To provide a realistic and
naturalistic scenario, the control schools were not encouraged to offer musical activities.
This measure was in consonance with the logical perspective of pragmatic RCTs which
may not employ placebos [16].
Music education is a complex intervention, mainly in an educational RCT context.
For example, it is impossible to standardize a day-by-day routine, as each class has a
different reality, and the music education might involve a huge spectrum of activities.
These activities include singing, exploration of rhythm (via corporal movement or
corporal percussion), and instrumental practice (which could be the highly technical
learning of a specific musical instrument, or using the instrument in an informal manner)
[15]. All of the procedures and activities described above are intended to: a) try to
systematize the same intervention based on the NCP, or b) try to provide the same
quality of intervention across various settings. Even with traditional educational methods
such as Kodaly (Hungarian method) or Orff (German method), day-by-day programs are
not established.
Description of Blinding
This RCT is an open label because the children who were selected for the
intervention knew that they were receiving music classes. At the same time, the
A r t i g o s | 83
selected intervention schools (and their scholar communities, i.e., teachers and
directors who were responsible to collect the secondary outcomes) knew about the
children who were allocated to receive intervention.
Statistical Analysis
Two different types of analyses were used to evaluate the effectiveness of the
music classes. The first (and standard) method was intention-to-treat (ITT), an approach
that assumes that every child in the intervention schools actually received the music
classes [28]. The other method, CACE estimation method took into account the
compliance status (children’s adherence to the music classes) [29,30]. The compliance
status is defined here as at least a 1% presence in the music classes during the five
months because with a presence of less than 1%, we are considering children who are
never-takers. CACE estimation, therefore, provides a realistic effect. Due to institutional,
organizational, and schedule differences (i.e., start and end of vacation period, holidays,
children’s regular examination period), the five intervention schools had different gross
numbers of musical classes (two schools had 57 musical classes, one 55, and another
50). Therefore, in order to take these differences into account in the CACE analysis, we
considered the percentage as a reference, instead of the gross number, to calculate the
compliance criteria.
Following the CACE estimation method, we have considered these assumptions:
1) the treatment assignment is random (as described above); 2) potential outcomes for
each child is unrelated to the treatment status of other individuals; 3) for never-takers
(children who do not receive the music classes even if they were assigned to this extra-
curricular activity) and always-takers, the distributions of the outcomes are independent
of the treatment assignment; 4) there are no defiers (children who do the opposite of
A r t i g o s | 84
what they are assigned to do); and 5) the average causal effect of the treatment
assignment on the treatment received is not equal to zero [31].
Although there is a practical issue motivating the using of cluster structure, a
statistical advantage exists in this design: It is very likely that individual interaction exists
between children from the same school conditions, which leaves the treatment condition
(control or intervention) less likely to be contaminated by other conditions. Therefore,
comparison of different conditions will be more valid [32].
The type of baseline distributions for the primary and secondary outcome
variables were considered (zero-inflated, normal, gamma). In addition, the standard
errors were adjusted for the survey design (i.e., taking the clusters into account), thus
generating robust standard errors (RSEs). Baseline significances tests comparing
children from control and intervention schools on its outcomes (primary and secondary)
and on potential confounders were conducted via t-Student or Mann-Whitney tests for
continuous outcomes (depending on its variance homogeneity and normality
distribution) and, for binary outcomes it was used the Chi-square.
Considering the CACE estimation method and ITT analysis, the primary
outcome was controlled by the confounders (visual acuity and central processing
assessment, IQ, and so on) along with age, gender, and baseline values from the same
outcome (i.e., word accuracy was controlled by word accuracy at baseline); the only
exception was adding the model involving phonological awareness as an outcome, as
visual acuity was not included.
A linear growth model was built for the ITT analysis of Portuguese and math
grades through the school year; for the CACE, linear growth mixture modeling was
used, allowing the incorporation of latent groups (complier and non-complier). Mplus
version 6.12 was used to build all regressions and general mixture models.
A r t i g o s | 85
Results
As suggested by CONSORT [28] the participants’ flow chart is described in the
Supporting Information and baseline measures comparing intervention and control
schools with its respective significances tests are described in Table 1.
As suggested by Assmann [33], in Table 1, we report a table of baseline data
with an overall description of the characteristics of the patients rather than using
significance tests. Although differences across groups at baseline were found, some
authors pointed out that the use of significance tests for detecting baseline differences
is questionable [34] and others that it is inappropriate [35,36]. Senn argued that “this
practice is philosophically unsound, of no practical value and potentially misleading”
[35].
Considering ITT, accuracy of word (β=2.57, p=0.047) has shown to be marginally
significant. This means children in the intervention school correctly read 2.57 words per
minute more than children in other schools do. Also in the ITT estimates, the slopes of
Portuguese (β=0.21, p=0.01) and math achievement (β=0.25, p<0.001) were
statistically significant for the intervention schools. This it means that every two
measured months, children from intervention schools increased 0.21 in Portuguese and
0.25 in math grades. There was no observed improvement in phonological awareness
(p=0.35) and in-text accuracy (p=0.23); non-word accuracy was negative and
nonsignificant (β=-1.512, p =0.40) (Table 2).
Regarding estimates for the complier group (using the CACE estimation method,
where comparisons are made considering the complier versus non-complier groups and
the effect of the control group is fixed at zero), estimates of word accuracy, in-text
accuracy, and phonological awareness are statistically significant; this means that
complier children read 13.98 more correct words per minute than children who are non-
A r t i g o s | 86
complier.. Indeed, positive slopes of Portuguese and math achievement showed to be
statistically significant.
Comparing the CACE and ITT, the CACE estimates were mostly higher than
those obtained using the ITT analysis (except for in-text accuracy and intercepts of
math and Portuguese). RSEs were lower in the CACE estimation method for the
primary outcome.
The ICC—the degree of correlation that is realized among outcomes of
participants in the same cluster—for each primary and secondary outcome, the pre- and
post-test results, and all respective standard errors and confidence intervals is shown in
Table 3. There was a considerable loss in statistical power for phonological awareness
variables due to an unexpectedly high degree of ICC variation. When the sample size
was estimated, low values were expected (approximately 0.025). The ICC confidence
intervals ranged from 0 to 0.596; the largest variation was observed in phonological
awareness (lower bound=0 and upper bound=0.596).
A positive growing slope (β=0.77, p=0.005) in Portuguese means that, every two
months, the grades in Portuguese increased 0.77 points for the complier group when
compared with non-complier children. Considering math (β=0.49, p<0.001), each two
months, the grades for the complier group increased 0.49, when compared with non-
complier group. The statistically significant and negative intercept indicates that the
Portuguese intercept for the complier group at baseline is 1.07 points lower than the
non-complier group; in math, the complier is 1.25 points lower.
Discussion and Considerations
The ITT analyses were rather unpromising: There was only one marginal
significant effect for the primary outcomes (accuracy of word reading) (p=0.047),
probably because if there is a real effect of music education, it could be attenuated
A r t i g o s | 87
among the children who were allocated to be in the intervention and have not taken it
(absence in the music classes, or presence of less than 1%) and the children who had
attained the music classes assiduously. However, taking into account complier status
via CACE estimation, it is possible to observe more promising effects in all primary
outcomes, in case of accuracy of word reading, it becomes 6 times bigger (from 2.57 to
13.98).
The only negative exception was for non-word accuracy, which was not
statistically significant by either the CACE or ITT estimation. This finding may have
resulted from the baseline rate of non-words per minute, which was superior in the
intervention group using the ITT method (Table 1). Although in-text accuracy with CACE
was lower than with ITT, it showed statistical significance for the former but not the
latter, corroborating to the idea that when we consider CACE estimation the effects of
intervention become more apparent
The negative estimations (CACE and ITT) for non-word accuracy are not
explained by the baseline differences between intervention and control schools
(significance test showed p-value=0.43) and were not significant at 0.05 for both
analyses (for ITT, p-value = 0.40, and for CACE, the p-value = 0.18). Indeed, a possible
interpretation for the unsettling negative value might be related to the automatized
process of word-level recognition, which was assimilated by children from intervention
schools (i.e., children from intervention schools performed better in word-accuracy
tasks). Maybe children were reading non-words as words (i.e., the more rapidly
automatized and more correct children read words, less precise they read the non-
words because children may read non-words as words). However, this hypothesis was
not our focus, and it might only be assessed via the evaluation of non-word reading task
errors’ typology.
A r t i g o s | 88
The positive slopes of Portuguese and math grades indicate that, throughout the
academic year, children from schools allocated to be in the intervention (general effect
via ITT analysis) and the complier children (CACE estimation) have trajectories that are
not flat, being music education effective for improvement of academic achievement.
For the secondary outcomes, there was a higher probability of estimating the
effects when considering a power of 0.8 because more than two assessments of
Portuguese and math grades were collected. This effect formed a third level in the
hierarchical model (first level formed by the child, second represented by the school,
and third by the four equi-distant measures in grades throughout the school year). The
ICC also was lower (ρ=0.06).
Table 3 described different magnitudes of ICCs, which may be interpreted as the
Pearson correlation coefficient between any two responses in the same cluster,
measuring the degree of similarity among responses within a cluster [37]. High (and
non-predicable) ICC values were obtained and directly influence our results; as a
consequence, our statistical power tends to be reduced in outcomes where ICCs were
inflated (i.e., in a general view, we underestimated the ρ value in the sample calculation
[ρ=0.025]). Values presented in Table 3 are important to guide future RTC research
involving scholar populations with measures related to learning and reading abilities.
However, the underlying reasons for variation between cluster will differ from trial to
trial, but two points in a cluster randomized study, particularly one involving education
strategies, might be addressed, as stated by Donner & Klar [38]: 1) the effect of
personal interaction among cluster members who received the same intervention; and
2) the influence of covariates at the cluster level, where all individuals in a cluster are
affected in a similar manner as a result of sharing exposure to a common environment.
A r t i g o s | 89
Some important limitations must be highlighted. First, due to issues which are
inherent to pragmatic RCTs, when ITT is estimated, control schools may not have an
active placebo (i.e., a “non-active” or placebo program was not introduced).
Consequently, part of the improvement in reading skills and in Portuguese and math
grades, in the ITT analysis, could result from the attention the music teachers paid to
the students. Various developmental antecedents (social deprivation, socioeconomic
status, family size, maternal reading, a stimulating home environment, maternal
depression, and child negligence) are small but significantly related to reading
achievement [38].
Because our children came from impoverished neighborhoods in Sao Paulo city,
they may be influenced by these non-measured developmental antecedent factors, and,
as a consequence, the musical activity may have functioned in two different ways: 1) as
a psychological effect due to the “extra” attention from music teachers, and 2) as an
environmental effect due to the provided stimulation itself (e.g., dance classes also
would provide perceptions of rhythm). Therefore, to argue that the development of
musical perception skills can account completely for the improvement in reading and
academic achievement would be misleading in this experiment. Furthermore, because
musical perception skills were not assessed throughout the full longitudinal study, we
cannot presume that the more musical skills, the better the improvement of reading
skills in our population will be. However, this pragmatic RCT did not aim to evaluate
what in music classes would improve reading and academic achievement, but to
pragmatically evaluate the effectiveness of music education as an intervention for
reading difficulties.
Considering estimates of CACE, considerations about placebo are irrelevant
because, as it was pointed out about the CACE assumptions, the effect of the control
A r t i g o s | 90
group is fixed at zero. The focus was exclusively on the complier and non-complier
groups that were compared with one another.
Lately, the reading measures also were limited to the decoding process and
methaphonological skills (word-level reading skills); therefore, we did not study reading
skills beyond word-level decoding, such as comprehension.
Future Implications
Based on the ICCs obtained in this study, future researchers should consider at
least 24 schools (12 intervention schools and 12 control schools) with 24 participating
children per school in order to reduce issues with power due to high variations in ICC,
as was observed with the phonological awareness variable. Increasing the number of
children per class does not significantly solve the power problem. At the same time, this
increase would likely make it more difficult for the music teachers to properly conduct
the musical activities. If some effect “exists,” the number of schools must be increased
in order to increase the degree of power in future research for outcomes with high ICC
variations. In future models and exploratory trials, placebo interventions (e.g., cooking
classes) also should be implemented, while measures related to developmental
antecedents should be evaluated and used as covariates.
Despite the noted limitations, this first RCT about music education is pragmatic
and showed promising positive effects on reading skills and academic achievement
considering CACE estimation, corroborating the theoretical rationale behind the music-
based intervention, which admittedly is an unorthodox approach (for details see [39]).
However, before recommending music classes as a public policy, more investigation
and data about the effectiveness of music education and theoretical models explaining
the impact of music abilities on reading skills are necessary, particularly in
A r t i g o s | 91
countries/scholar populations with low estimates of reading performance and academic
achievement, as well as high levels of disparity between public and private schools.
.
A r t i g o s | 92
References
1. Cunningham AE, Stanovich KE (2001) What reading does for the mind. Journal of Direct Instruction 1: 137-149.
2. Snow CE, Burns MS, Griffin P (1998) Preventing reading difficulties in young children. Washington, DC: National Academies Press.
3. Torgesen JK (2000) Individual differences in response to early interventions in reading: The lingering problem of treatment resisters. Learn Disabil Res Pract 15: 55-64.
4. Duff F, Clarke P (2011) Practitioner Review: Reading disorders - What are effective interventions and how should they be implemented and evaluated? J Child Psychol Psychiatry 52: 3-12.
5. Kolinsky R, Cuvelier H, Goetry V, Peretz I, Morais J (2009) Music Training Facilitates Lexical Stress Processing. Music Percept 26: 235-246.
6. Moreno S, Marques C, Santos A, Santos M, Castro SL, et al. (2009) Musical training influences linguistic abilities in 8-year-old children: More evidence for brain plasticity. Cereb Cortex 19: 712-723.
7. Foxton JM, Talcott JB, Witton C, Brace H, McIntyre F, et al. (2003) Reading skills are related to global, but not local, acoustic pattern perception. Nat Neurosci 6: 343-344.
8. Pantev C, Oostenveld R, Engelien A, Ross B, Roberts LE, et al. (1998) Increased auditory cortical representation in musicians. Nature 392: 811-814.
9. Anvari S, Trainor L, Woodside J, Levy B (2002) Relations among musical skills, phonological processing, and early reading ability in preschool children. J Exp Child Psychol 83: 111-130.
10. Patel A, Iversen J (2007) The linguistic benefits of musical abilities. Trends Cogn Sci 11: 369-372.
11. Barnett S, Ceci S (2002) When and where do we apply what we learn? A taxonomy for far transfer. Psychol Bull 128: 612-637.
12. Besson M, Schon D, Moreno S, Santos A, Magne C (2007) Influence of musical expertise and musical training on pitch processing in music and language. Restor Neurol Neurosci 25: 399-410.
13. Patel AD (2008) Science & music: talk of the tone. Nature 453: 726-727. 14. Zatorre R, McGill J (2005) Music, the food of neuroscience? Nature 434: 312-315. 15. Cogo-Moreira H, Andriolo RB, Yazigi L, Ploubidis G, Brandão de Ávila CR, et al.
(2012) Music education for improving reading skills in children and adolescents with dyslexia. Cochrane database sys rev 15: CD009133.
16. Hotopf M (2002) The pragmatic randomised controlled trial. Advances in Psychiatric Treatment 326-333.
17. Cogo-Moreira H, Ploubidis G, de Avila C, Mari J, Pinheiro A (2012) EACOL (Scale of Evaluation of Reading Competency by the Teacher): Evidence of Concurrent and Divergent Validity. Neuropsychiatr Dis Treat 8: 443-454.
18. Pasquali L, Wechsler S, Bensusan E (2002) Raven's colored progressive matrices for children: a validation study for Brazil. [Raven´s Colored Progressive Matrices for Children: a validation study for Brazil]. Aval Psicol 1: 95-110.
19. Pereira L (1993) Processamento auditivo. [Hearing Processing]Temas desenv 12: 7-14.
20. Cruz MBZ (2005) WISC III: Escala de Inteligência Wechsler para crianças: Manual. Aval Psicol 4: 199-201.
A r t i g o s | 93
21. Wechsler D (1991)WISC-III: Wechsler intelligence scale for children. San Antonio, TX: The Psychological Corporation.
22. Capovilla A, Capovilla F (1998) Prova de consciência fonológica: desenvolvimento de dez habilidades da pré-escola à segunda série [Phonological Awareness Test: development of tem abilities from presschool to third grade].Temas desenvol 7: 14-20.
23. Hulme C, Hatcher PJ, Nation K, Brown A, Adams J, et al. (2002) Phoneme awareness is a better predictor of early reading skill than onset-rime awareness. J Exp Child Psychol 82: 2-28.
24. Bradley L, Bryant PE (1978) Difficulties in auditory organisation as a possible cause of reading backwardness. Nature 271: 746-747.
25. Snowling M, Hulme C, Nation K (1997) A connectionist perspective on the development of reading skills in children. Trends Cogn Sci 1: 88-91.
26. Brazil (1997) Introdução aos parâmetros curriculares nacionais In: Ministry of Education. Brasilia.
27. NafME National Association for Music Education (2012). Avaliable: www.nafme.org. Acessed: 12 January 2013.
28. Campbell M, Elbourne D, Altman D (2004) CONSORT statement: Extension to cluster randomised trials. BMJ 328: 702-708.
29. Angrist JD, Imbens GW, Rubin DB (1996) Identification of causal effects using instrumental variables. J Am Stat Assoc 91: 444-455.
30. Jo B, Muthén BO (2001) Modeling of intervention effects with noncompliance: a latent variable approach for randomized trials. New developments and techniques in structural equation modeling: 57-87.
31. Jo B (2002) Model misspecification sensitivity analysis in estimating causal effects of interventions with non-compliance. Stat Med 21: 3161-3181.
32. Sobel M (2006) What Do Randomized Studies of Housing Mobility Demonstrate? . J Am Stat Assoc 101: 1398-1407.
33. Assmann S, Pocock S, Enos L, LE K (2000) Subgroup analysis and other (mis)uses of baseline data in clinical trials. Lancet 355: 1064-1069.
34. Altman D, Dore CJ (1990) Randomisation and baseline comparisons in clinical trials. Lancet 335: 149-153.
35. Altman D (1985) Comparability of randomised groups. Statistician 34: 125-136. 36. Senn S (1994) Testing for baseline balance in clinical trial Stat Med 13: 1715-1726. 37. Donner A, Klar N (2000) Design and Analysis of Cluster Randomization Trials in
Health Research. West Sussex: John Wiley & Sons Ltd. 178 p. 38. Trzesniewski KH, Moffitt TE, Caspi A, Taylor A, Maughan B (2006) Revisiting the
association between reading achievement and antisocial behavior: New evidence of an environmental explanation from a twin study. Child Dev 77: 72-88.
39. Snowling M, Hulme C (2011) Evidence-based interventions for Reading and language difficulties: creating virtuous circle. Br J Educ Psychol 81: 1-23.
A r t i g o s | 94
Table 1 – Comparisons (absolute values or means with theirs Robust Standard Errors)
between control and intervention and its respective significance test.
Variables at baseline Intervention
Schools
Control
Schools
p-value
Number of schools 5 5
Number the children 114 121
Number of children
(max) per school
27 27
Number of children
(min) per school
17 23
Mean of Accuracy of
word (RSE)
9.44(0.82) 11.22(3.60) 0.79
Mean of Accuracy of
non-word (RSE)
5.90(1.10) 5.16(1.42) 0.43
Mean of Phonological
Awareness (RSE)
25.78(0.70) 23.95(0.70) <0.001
Mean of Portuguese
(mean and RSE)
4.35(0.10) 5.33(0.23) <0.001
Mean of Math (RSE) 4.49(0.19) 5.5(0.27) <0.001
Mean of IQ (RSE) 91.30 (3.35) 93.43(2.38) 0.88
Children with problems
in SPTA
58 31 <0.001
Children with Visual
Acuity Problems
47 33 0.02
Mean of number the
children per class (RSE)
30.63(1.72) 31.11(2.46) 0.89
Attendance through
scholar year (RSE)
188.08(1.38) 188.25(1.77) <0.001
Drop out in the follow up 7 6
A r t i g o s | 95
Abbreviations: max=maximum; min=minimum; RSE=robust standard error
Table 2 -Effects of Music Education considering ITT and CACE
Intention-to-treat (intervention vs. control)
Complier-Average Causal Effect (complier vs. non-complier)
Outcomes Estimates RSE Estimates RSE
(Reading) Accuracy of non-word -1.39 1.67 -1.3 0.959
Accuracy of word 2.57* 1.29 13.983*** 0.853
Accuracy of text 3.00 2.519 0.41*** 2.412
Phonological Awareness 0.88 0.94 19.719*** 1.00
(Portuguese Achievement) Slope† 0.21** 0.076 0.77** 0.27
Intercep -1.00*** 0.311 -1.07*** 0.31
(Math Achievement) Slope † 0.246*** 0.062 0.491** 0.174
Intercept -0.004*** 0.344 -1.253*** 0.349
p-values are expressed as following: <=0.05(*),<=0.01(**), <=0.001(***)
†Slope on school status (i.e., on intervention schools, in case of ITT analysis) and slope
on CACE parameter (complier children, in case of CACE estimation)
RSE= Robust Standard Error
A r t i g o s | 96
Table 3 – Intraclass correlation coefficient for primary and secondary outcomes at
baseline and last assessment.
Outcomes Pre-test Post-test
ICC SE 95% CI ICC SE 95%CI
Primary Outcome Accuracy of word 0.18 0.08 0.01 0.34 0.22 0.1 0.03 0,41
Accuracy of non-word 0.11 0.06 0 0.23 0.24 0.1 0.04 0,43
Accurracy of text 0.12 0.07 0 0.26 0.15 0.08 0 0,30
Phonological awareness 0.14 0.07 0 0.29 0.36 0.12 0.12 0,60
Secondary Outcome
Portuguese (baseline and forth assessment) 0.06 0.05 0 0.16 0 0.02 0 0,04
Math (Baseline and fourth assessment) 0.2 0.09 0.02 0.37 0.75 0.05 0 0,18
Abbreviation: CI= Confidence Interval; SE= Standard Error; ICC= Intraclass Coefficient
Correlation
A r t i g o s | 97
3.4 Music Perception Predicts Word-level Reading Ability in Children with
Reading Difficulties
Under Review in Neuropsychology
Abstract
Objective: To investigate whether specific domains of musical perception
(temporal and melodic domains) predict the word-level reading skills of eight- to ten-
year-old children (n = 235) with reading difficulties, normal quotient of intelligence, and
no previous exposure to music education classes. Method: A general-specific solution
of Montreal Battery of Evaluation of Amusia (MBEA), which underlies a music
perception construct and is constituted by three latent factors (the general, temporal and
the melodic domain), was regressed on word-level reading skills (rate of correct isolated
words/non-words read per minute). Results: General part and the melodic domain
predicted word-level reading. Conclusions: This finding indicated that a) musical
perception has two specific domains and b) the phenomenon of melodic domain as a
predictor of word-level reading skills is found not only in children with dyslexia but also
children with reading difficulties.
Keywords: musical perception; reading difficulties; decoding; structural equation
modeling
A r t i g o s | 98
Introduction
Associations between musical ability and language skills have been studied over
the past twenty years. Studies have found that musical ability may facilitate the
acquisition of a second language, as it predicts phonological ability (both receptive and
productive) even when controlling for other factors (Slevc & Miyake, 2006). Musical
metrical perception, as one domain of musical perception, has an organisational
function in the phonology of language and phonological learning via speech prosody.
For example, it enables the accurate segmentation of syllables and words from the
speech stream (Echols, 1996), which is likely critical for phonological development and,
consequently, the development of literacy (Huss, Verney, Fosker, Mead, & Goswami,
2011).
Musical training has been found to facilitate the processing of lexical stress
(Kolinsky, Lidji, Peretz, Besson, & Morais, 2009), verbal memory (Ho, Cheung, & Chan,
2003), and verbal intelligence and executive function (Moreno et al., 2011).
On the basis of these findings, it has been hypothesised that musical intervention
for children with reading difficulties would be helpful because music aptitude and literacy
are both related to the extent of subcortical adaptation to regularities in ongoing speech
and to auditory working memory and attention. Therefore, similar brain mechanisms
underlie reading and music abilities (Strait, Hornickel, & Kraus, 2011). However, it is
important to note that music notation is a spatial and temporal representation of pitch
patterns. By contrast, the alphabet is an arbitrary representation of sound elements that
are constrained by the phonotactic patterns in language.
Anvari, Trainor, Woodside, and Levy (2002) have indicated that there is little
agreement on the particular elements of music perception that may correlate with
reading difficulties among school-aged children. Auditory analysis skills used in the
A r t i g o s | 99
processing of language, such as blending and segmenting sounds, are similar to the
skills necessary for music perception, such as rhythmic, melodic, and harmonic
discrimination, despite the differences in the nature of the stimulus (one musical and the
other linguistic) (Lamb & Gregory, 1993). Therefore, it is reasonable to hypothesise that
if early reading skills are closely related to skill in processing the auditory components of
speech, then the different elements of music perception (timbre, rhythm, harmony, and
pitch) may also be associated with reading development (Douglas & Willatts, 1994).
A systematic review was conducted to evaluate the effectiveness of music
education on reading skills. The review employed a sensitive search using dyslexia (a
specific learning disability) and general descriptive terms such as reading difficulties and
reading problems to identify randomised clinical trials in which music education was
used to improve the reading skills of children and adolescents. Among more than 700
citations without language restrictions, no randomised, controlled clinical trials were
published until June of 2012 (Cogo-Moreira, Andriolo, et al., 2012). However, recently, a
pragmatic cluster randomised clinical trial conducted in ten public schools in São Paulo
found promising results of music education (intervention was offered over 5 months, 3
times per week, 50 minutes per class) among children from eight to ten years old with
reading difficulties. Specifically, considering the causal-average complier effect’s
analysis, children’s word-level reading skills (decoding and phonological awareness)
and, secondarily, academic achievement in Portuguese and math improved (Cogo-
Moreira, Brandao de Avila, Ploubidis, & Mari Jde, 2013). The explanation for the causal
paths to reading development via musical training may be referred to as
“transfer” (Barnett & Ceci, 2002; Patel, 2008). The connection between musical learning
and improved reading skills is a “far transfer” because musical learning is not directly
related to reading or academic achievement in Portuguese or math. Musical training is
A r t i g o s | 100
based on teaching and constant practice of non-verbal structures such as classical
sheet music and contact with elements that are present in musical phenomena (e.g.,
rhythm, melody, meter, and so on), whereas reading is related to verbal language. An
example of “near transfer” is the acceleration of the development of harmony perception
in 4- and 5-year-olds due to musical training (Corrigall & Trainor, 2009).
An important question to ask about the promising results of music education (i.e.,
musical learning) on reading skills in the first (above-cited) randomised clinical trial of
this intervention is whether the improvement in reading and academic achievement was
caused by the improvement in musical perception. The authors only collected measures
related to musical perception skills at baseline. Therefore, it was not possible to
determine how and why music education (and, thus, a potential improvement in musical
perception skill) might improve reading and academic improvement. The pragmatic
randomised clinical trial was designed, as the name indicates, to reflect the
heterogeneity of children with reading difficulties. It was not concerned with narrow
diagnostic labels (Hotopf, Churchill, & Lewis, 1999) and, therefore, focused on
effectiveness (not efficacy) of music education for word-level reading skills.
The purpose of the current study was to test whether music perceptions and their
specific latent domains predict word-level reading skills (e.g., decoding of word and non-
word) using a structural equation model with latent factors.
Methods
This study was conducted using the baseline measurements and sample from a
randomised clinical trial of the effectiveness of music education for improving the
reading skills and academic achievement of children with poor reading skills (Cogo-
A r t i g o s | 101
Moreira et al., 2013). Briefly, the sample comprised 235 children from eight to ten years
old (girls = 38.3%), with an average age of 9.15 years (SD = .05), with reading
difficulties from ten public schools in impoverished neighbourhoods on the outskirts of
the city of São Paulo. The 48 teachers from the ten schools were asked to complete the
Scale of Assessment of Reading Competence of Students by the Teacher (EACOL),
following the instruction: “…for the children in your class with reading ability below the
mean for the corresponding grade, please fill out the EACOL.” The EACOL contains 27
dichotomous items that evaluate the following two domains of elementary children’s
reading competences: reading aloud (17 items; related mostly to automatised word
recognition and decoding skills) and silent reading ability (10 items; related mostly to
comprehension). The EACOL distinguishes three latent groups of readers (good reader,
not-so-good reader, and poor reader), showing good discriminant and concurrent
validity (Cogo-Moreira, Ploubidis, de Avila, Mari, & Pinheiro, 2012). Scores can range
from -29 to +29 points and are interpreted as follows: poor reader (< -14.5), not-so-good
reader (from 14.5 to 14.5), and good reader (> 14.5). Of the 235 children, 82.80% were
classified as poor readers and 17.2% as not-so-good readers.
Lastly, other important features of this sample are as follows:
these children’s non-verbal intellectual ability, measured by Raven’s
coloured Progressive Matrices (Pasquali, Wechsler, & Bensusan, 2002),
was above the 25th percentile.
they did not receive any regular language-speech therapy or music
classes (such as private music classes in a conservatory or other school
of music, e.g., in a social project involving musical learning).
Information on these two features were obtained by the parents, who provided
consent for their children to participate in this study.
A r t i g o s | 102
The variables used to build a model to test whether music perception and its
specific latent domains predict word-level reading skills are described below.
Materials
Potential confounders
The Simplified Auditory Processing Test (SAPT) (Pereira, 1993)
was conducted by a hearing and speech pathologist. The following auditory
abilities were tested: sound localisation in five directions and verbal and non-
verbal sequential memory, corresponding to the processes of localisation and
temporal time ordering. The children were classified as having or not having
problems in central auditory processing.
For Intelligence Quotient (IQ), the complete Brazilian Portuguese
validation of the Wechsler Intelligence Scale for Children-Third Edition (WISC-III)
was used and administered by a trained psychologist (Figueiredo, 2002;
Wechsler, 1991).
To measure the ability to analyse metaphonological skills, we used
the Test of Phonological Awareness (Capovilla & Capovilla, 1998). The test is
composed of ten subtests, each consisting of four items that test the following:
synthesis, segmentation, syllabic and phonemic manipulation and transposition,
and rhyme and alliteration.
Visual acuity of the children (age-appropriate) under conditions of
monocular viewing was tested by an ophthalmology technician using Snellen’s
chart. The children were classified as either having visual alterations or not.
Word-level reading skills assessments
Decoding
A r t i g o s | 103
Variables related to word-level reading skills were collected to build the model.
These skills included the following:
accuracy of word task (rate of correct words read per minute)
accuracy of non-word task (rate of correct non-words read per minute)
The lists used to assess skills had a total of 88 words and 88 non-words. The
lists contained both high- and low-frequency words, words with bi-directional regularity
(regular and irregular words according to grapheme-phoneme/phoneme-grapheme
correspondence), and words of various lengths (short, medium, and long words in terms
of the number of letters). The non-words were built with identical orthographic Brazilian
Portuguese structure and had the identical length of stimuli used in the list of words.
Children were asked to read aloud the words and non-words, and the time spent
reading was computed. Segmentation and prolongation in the children’s reading were
considered faults.
The correlation between word and non-word accuracy tasks was high (r =.92, p <
.001); the tasks showed a moderately positive correlation with the Phonological
Awareness Test (Capovilla & Capovilla, 1998) (r accuracy of word = .40, p < .001 and r accuracy
of non-word = .37, p < .001). As expected, general Intelligence Quotient (IQ) was poorly
related to word accuracy (r = .168, p = .01) and not correlated with non-word accuracy (r
= .01, p = .131). For more detailed information about the validity of word and non-word
tasks, see (Cogo-Moreira, Ploubidis, et al., 2012).
Fluency in Text
Accuracy of text (rate of correctly read words per minute) was obtained with
consideration for the age of the child. For the eight-year-old children, we used The
Tortoise and the Leopard (Barros, 1995); for the nine-year-olds, we used The Nut
Veterinarian (Pereira, 1981b); and for the 10-year-olds, we used The Owl and the Eagle
A r t i g o s | 104
(Pereira, 1981a). The children’s reading was recorded for post-analysis of accuracy.
The baseline assessments of accuracy of text (rate of correct words read per minute)
was highly correlated with the accuracy of words (r = .916; p < .001) and the accuracy
of non-words (r = .873; p < .001).
Music perception
The Montreal Battery for Evaluation of Amusia (MBEA), which consists of six
subtests (Scale, Contour, Interval, Rhythm, Meter, and Melody Memory), was used to
evaluate the mechanisms that underlie music perception. The MBEA does not rely on
the memory of familiar melodies and was developed considering the cognitive theories
of music perception and neuropsychological evidence (Peretz & Coltheart, 2003).
Children are given 10 minutes to work on each of the six subtests. The subtests are
composed of 30 to 31 trials, which are preceded by at least two examples, with
feedback and question-answering allowed. No feedback is provided during the test. The
participants listened to two melodies and indicated whether these melodies were the
same or different. The Portuguese version of the MBEA was validated for the Brazilian
social and cultural context (Nunes-Silva & Haase, 2012). For this study, the melody
memory subtest (the last MBEA subtest) was not used because the majority of children
asked to stop the examination during the first days of baseline assessment due to
fatigue. Therefore, for ethical reasons, only five subtests from the MBEA were used.
Statistical Analysis
Confirmatory factor analysis (CFA) of the MBEA: The general-specific model
Peretz, Champod, and Hyde (2003) and Cooper, Tobey, and Loizou (2008);
(Peretz et al., 2003) presented a musical perception model, in which the MBEA’s
subtests were separated into pitch-based tests (which is called here melodic domain)
A r t i g o s | 105
and temporal-based tests. Using CFA, we tested the fit of this musical perception model
among the current study’s young population with reading difficulties
The correlations between the general-specific factor (also called bi-factor model)
and the specific factors (i.e., melodic and temporal domains) and between the specific
factors and the general factor (i.e., musical perception) were fixed at zero. The general
factor is loaded (explained by) all five MBEA subtests; specific refers to two latent
factors (melodic and temporal) that account for the association between music
perception indicators and the specific dimensions/factors. The Comparative Fit Index
(CFI), the Tucker Lewis Index (TLI), and the Root Mean Square Error of Approximation
(RMSEA) were used to evaluate the model’s fit. The CFI refers to the discrepancy
function adjusted for sample size. The TLI assesses the incremental fit of a model
compared to a null model. Both range from 0 to 1, and an acceptable model fit is
indicated by a CFI and TLI value of .95 or greater. The RMSEA is related to residuals in
the model ranging from 0 to 1, and an acceptable model fit is indicated by an RMSEA
value of .06 or less. The Standardised Root Mean Square Residual (SRMR) is an
absolute measure of fit and is defined as the standardised difference between the
predicted correlation and the observed correlation; a value less than .08 is generally
considered a good fit. These indices of fit and a chi-squared goodness of fit test were
used to assess the model fit, as suggested by existing guidelines (Hu & Bentler, 1999).
The CFA, as a first step in developing an integrative model, was utilised to create
latent domains underlying MBEA’s music perception construct. Latent domains present
opportunities to understand how temporal and melodic domains (as latent domains)
could affect word-level reading skills better than each of the MBEA’s subtests
separately or simply by adding the scores from the subtests.
4.2. Integrative Model—Reading Skills and Music Perception Abilities
A r t i g o s | 106
The general-specific factor model of musical perception (based on theoretical
and statistical considerations) was regressed on reading variables and confounders.
The rationale for the causality concatenation among the variables in the integrative
model was based on the following assumptions: (a) music perception influences word-
level reading skills (i.e., word and non-word accuracy); (b) these skills influence fluency
in text (i.e., rate of correct words read per minute); and (c) confounders impact fluency
in text. Therefore, there is concatenation beginning with music perception, passing
through word-level skills to fluency in text.
The accuracies of words and non-words were treated as censored from below
(that is, having a floor effect) because an excess-zero issue may cause significant
correlation coefficient results and false correlation findings in the proposed structural
equation. Because the sample consisted of children with reading difficulties, some
children scored zero correct read words and non-words per minute, generating floor
effects in both word-level reading measures. The maximum likelihood estimator was
used to perform the integrative model (reading variables, confounders, and musical
perception). To account for the non-independence of observations (i.e., children nested
in schools, which generated a multilevel structure), standard errors were computed
based on (Asparouhov & Muthén, 2005, 2006). Likely problems related to non-
convergence due to negative residual variances and correlations greater than one
caused by dependence in the data set were inspected. All continuous variables and
covariates were tested for univariate normality based on skewness and kurtosis tests.
The structural equation modelling (CFA and path analysis) was conducted via
Mplus 7 (Muthén & Muthén, 1998-2012).
Results
Sample description
A r t i g o s | 107
The current sample included 235 children (38.29% girls), of whom 37.87% had
problems in auditory processing and 34.04% had visual acuity problems. The average
age was 9.15 (SD = .05). The mean number of children in each school was 23. Table 1
displays means and standard deviations (SD), kurtosis, skewness, and univariate
normality test (skewness and kurtosis normality test) of reading measures and other
collected continuous variables.
Table 1. Mean, standard deviation (SD), kurtosis, skewness, and univariate normality
test (skewness and kurtosis normality test) for all continuous (observable and latent)
variables
Variables Mean SD Kurtosis Skewness S-K normality test
(p-value)
Age 9.19 0.81 2.27 0.28 <0.001
IQ total 92.40 13.92 2.89 0.13 0.68
Phonological awareness 24.88 5.12 3.44 -0.17 0.2
Accuracy of word 10.34 12.95 4.91 1.53 <0.001
Accuracy of non-word 5.53 7.44 5.44 1.72 <0.001
In-text accuracy 26.17 21.36 3.76 1.03 <0.001
General music perception
factor score 0.00* 1.09 2.63 0.30 0.08
Melodic factor score 0.00* 0.74 3.31 0.19 0.25
Temporal factor score 0.00* 1.60 3.11 -0.11 0.67
Abbreviations: SD = standard deviation; S-K = skewness and kurtosis
A r t i g o s | 108
*Differences in the mean of factor scores are observable only in the fifth value following
the period. To avoid scientific notion (i.e., 4E-5), we opted present the mean of factor
loadings as zero.
The musical perception model fit
We ran the musical perception model isolated from the other measures; two
children were absent at the time of evaluation. Three models of the MBEA’s underlying construct were tested
via CFA, as follows: a general-specific model, a second-order model, and a general model.
The general-specific model displayed the following indices: CFI = 1.0, TLI =
1.019, RMSEA = 0, and SRMR = .012. For each domain, factor scores were obtained
and were based on the regression methods (also known as the maximum posterior
estimator, see Lawley and Maxwell (1962)). To avoid the sum of subtests, we computed
the factor score, resulting in a measure of latent domains. The factor scores for each of
the domains were normally distributed, with a mean at zero and standard deviation
close to one (Table 1).
Musical perception, reading, and confounders–an integrative model
The first integrative model, in which the general-specific model was regressed on
word and non-word accuracy and confounders, is shown in Figure 1 (dashed and solid
lines and squares).
A r t i g o s | 109
Figure 1 – Integrative model with convergence problems
The model returned a non-positive residual covariance matrix. No negative
variance, residual variance, or even correlation of one or greater was observed.
Dependency among some variables may have caused the convergence problems due
to a high correlation among the variables. Specifically, word and non-word accuracy
may have generated this convergence problem and, therefore, non-word reading was
excluded from the model. Because a convergence problem is an inadmissible solution,
we present the diagram only to provide the hypothesised model, but related
standardised coefficients (and their statistical significances) are not presented. We
excluded the non-reading accuracy route (dashed lines) from the model because when
A r t i g o s | 110
children develop reading fluency, they abandon the phonological strategy, which is
more related to non-word reading skills, and rely on the graphemic encoding procedure.
Due to the non-positive matrix, a second model was tested (Figure 2), in which
the general factor of musical perception (b = 2.57, p = .014) and a specific part (the
melodic domain [b = 2.105, p = .035]) were predictors of word accuracy. Thick lines and
letters in Figure 2 indicate statistically significant trajectories, and presented values are
the standardised coefficients.
Figure 2 – Integrative model with standardised coefficients
Regarding the covariates, phonological awareness (b = 1.318, p < .001) was a
statistically significant predictor of in-text accuracy (fluency in text). All of the above
regression coefficients values are non-standardised (called “b”). In Figure 2, they are
standardised.
A r t i g o s | 111
Discussion
A high correlation between word and non-word accuracy (r = .92, p < .001)
suggested that both decoding measures tapped the same construct rather than two
distinguishable constructs. Indeed, because the children in this study had weak word
reading skills, the word reading task may have, in effect, been a non-word reading task
for them. This high correlation motivated the omission of non-word accuracy in the
integrative model (dashed square and its dashed trajectories) to avoid convergence
problems.
Melodic domain was a predictor of word accuracy (model with solid lines). This
dissociation involving a specific domain of music perception may be related to the
independent processing of musical (melodic and temporal) domains. The traditional
neuropsychological view has treated them separately (Peretz & Zatorre, 2005);
however, other authors have argued that perception, attention, and memory for pitch
relations are inherently rhythmical because the perception of melody and rhythm is
treated as a unified dimension (Jones & Boltz, 1989).
The decision regarding the underlying structure of the MBEA as a general-
specific model was theoretical, as hypothesised by (Peretz et al., 2003), corroborating
descriptions about two parallel and independent subsystems (Carroll-Phelan &
Hampson, 1996).
The melodic domain’s (a specific part of the music perception model) prediction
of word accuracy may be related to the phonological skills that are required in word-
level reading tasks. One study found that global pitch processing (related to the melodic
domain) and reading component skills were restricted to the phonological domain,
establishing a semi-partial correlation between global pitch perception tasks and non-
A r t i g o s | 112
word repetition tasks (r = .38; p < .05) and speed in reading aloud a list of non-words (r
= .41; p < .05) (Foxton et al., 2003).
Because the children in the current study had difficulties in reading, as previously
mentioned, the list of words may have served as non-words, generating convergence
problems when the accuracy of words and non-words were used in the same model.
The present results were consistent with previous findings, especially with
studies involving children with dyslexia; however, it is important to emphasise that the
current sample was not given such a diagnosis (dyslexia). Therefore, the results may
indicate the following: (1) achievement on pitch-based tasks (as a predictor) is not
exclusive to children with dyslexia; it also occurs in children with reading difficulties
(remembering that only word-level decoding skills were evaluated in the present study)
and (2) a possible explanation of how and where musical learning might improve the
development of word-level reading skills. However, there is a methodological limitation
in the later perspective. It is based on a cross-sectional point of view because we only
considered baseline measurements in a randomised clinical trial via structural equation
modelling. Consequently, causality cannot be determined. A study involving children
aged eight to eleven years old found a link between impaired pitch processing and
abnormal phonological development in children with dyslexia, demonstrating that pitch
pattern processing is an important predictor for the early diagnosis and remediation of
dyslexia (Ziegler, Pech-Georgel, George, & Foxton, 2012). In contrast with these
findings, (Degé & Schwarzer, 2011) have shown that enhancement of phonological
awareness among pre-schoolers was driven by more general positive effects of the
music program. Dellatolas, Watier, Le Normand, Lubart, and Chevrie-Muller (2009)
have suggested that for children with dyslexia, the link between written language and
A r t i g o s | 113
music perception is at the rhythmic level, not at the level of pitch, as the simple rhythm
reproduction tasks in kindergarten are predictive of later reading performance.
Regarding the music perception evaluation, it is important to stress that MBEA is
based on the paradigm of Western European music of the tonal period (more
specifically, it considers the majority of the items to have a diatonic structure). Other
paradigms underlying a music perception model involving atonal structure and other
spectra of metrics (e.g., 7:8, 12:15) and timbres (white or pink noise) were not
considered but might be examined in future studies involving predictions of music
perception and its relation to reading skills. Additionally, the current study’s reading
variables represented level of decoding and, as a consequence, other reading skills
(higher level language skills related to comprehension). There is a strong body of
evidence demonstrating that phonological awareness is a predictor of reading
development and spelling (Hulme, Bowyer-Crane, Carroll, Duff, & Snowling, 2012).
Future randomised clinical trials involving children might provide additional
evidence of (1) the reading-related benefits of music education that might be replicable
in other contexts and cultures (languages), especially showing the “transfer” to higher-
level reading skills, for example, comprehension and (2) reading and academic
achievement progress that might correspond to improvements in music perception. The
thorough examination of these issues could contribute to the formation of public policy
involving music education implementation in curricula. If music education can improve
music perception skills (which was not shown in the first pragmatic randomised clinical
trial of Cogo-Moreira et al. (2013) and, in turn, help children with reading disabilities
(e.g., assisting in the processing of lexical skills (Kolinsky et al., 2009) and improving in
both speech and reading (Moreno et al., 2009), then countries with low achievement in
A r t i g o s | 114
reading (e.g., Peru, Panama, Montenegro, Bulgaria, and the Russian Federation) may
benefit.
Conclusion
As a main conclusion, the MBEA’s melodic domain was a predictor of word-level
decoding skills among children with reading difficulties, corroborating evidence of the
dissociation of musical perception in the temporal and melodic domains.
References
Anvari, S. H., Trainor, L. J., Woodside, J., & Levy, B. A. (2002). Relations among
musical skills, phonological processing, and early reading ability in preschool children. J Exp Child Psychol, 83(2), 111-130.
Asparouhov, T., & Muthén, B. (2005). Multivariate statistical modeling with survey data. Paper presented at the Proceedings of the FCMS.
Asparouhov, T., & Muthén, B. (2006). Multilevel modeling of complex survey data. Paper presented at the Proceedings of the Joint Statistical Meeting Seatle.
Barnett, S. M., & Ceci, S. J. (2002). When and where do we apply what we learn? A taxonomy for far transfer. Psychol Bull, 128(4), 612-637.
Barros, M. (1995). O jabuti e a onça. Sao Paulo: Editora do Brasil. Capovilla, A. G. S., & Capovilla, F. C. (1998). Prova de consciência fonológica:
desenvolvimento de dez habilidades da pré-escola à segunda série. Temas sobre desenvolvimento, 7(37), 14-20.
Carroll-Phelan, B., & Hampson, P. J. (1996). Multiple components of the perception of musical sequences: A cognitive neuroscience analysis and some implications for auditory imagery. Music Perception, 13(4), 517-561.
Cogo-Moreira, H., Andriolo, R. B., Yazigi, L., Ploubidis, G. B., Brandao de Avila, C. R., & Mari, J. J. (2012). Music education for improving reading skills in children and adolescents with dyslexia. Cochrane Database Syst Rev, 8, CD009133. doi: 10.1002/14651858.CD009133.pub2
Cogo-Moreira, H., Brandao de Avila, C. R., Ploubidis, G. B., & Mari Jde, J. (2013).
Effectiveness of music education for the improvement of reading skills and academic achievement in young poor readers: a pragmatic cluster-randomized, controlled clinical trial. PLoS One, 8(3), e59984. doi: 10.1371/journal.pone.0059984
Cogo-Moreira, H., Ploubidis, G. B., de Avila, C. R., Mari, J. J., & Pinheiro, A. M. (2012).
EACOL (Scale of Evaluation of Reading Competence by the Teacher): evidence
A r t i g o s | 115
of concurrent and discriminant validity. Neuropsychiatric disease and treatment, 8, 443-454. doi: 10.2147/NDT.S36196
Cooper, W. B., Tobey, E., & Loizou, P. C. (2008). Music perception by cochlear implant
and normal hearing listeners as measured by the Montreal Battery for Evaluation of Amusia. Ear Hear, 29(4), 618-626. doi: 10.1097/AUD.0b013e318174e787
Corrigall, K. A., & Trainor, L. J. (2009). Effects of musical training on key and harmony
perception. Annals of the New York Academy of Sciences, 1169, 164-168. doi: 10.1111/j.1749-6632.2009.04769.x
Degé, F., & Schwarzer, G. (2011). The effect of a music program on phonological
awareness in preschoolers. Frontiers in psychology, 2(124). doi: 10.3389/fpsyg.2011.00124
Dellatolas, G., Watier, L., Le Normand, M. T., Lubart, T., & Chevrie-Muller, C. (2009).
Rhythm reproduction in kindergarten, reading performance at second grade, and developmental dyslexia theories. Archives of clinical neuropsychology 24(6), 555-563. doi: 10.1093/arclin/acp044
Douglas, S., & Willatts, P. (1994). The relationship between musical ability and literacy
skills. Journal of Research in Reading, 17(2), 99-107. Echols, C. (1996). A role for stress in early speech segmentation. In J. Morgan & K.
Demuth (Eds.), Signal to Syntax: Bootstrapping from Speech to Grammar in Early Acquisition (pp. 151-170). Mahwah, NJ: Lawrence.
Figueiredo, V. (2002). WISC-III: Escala de Inteligência Wechsler para Crianças-
adaptação brasileira da 3ª edição. São Paulo: Casa do Psicólogo. Ho, Y. C., Cheung, M. C., & Chan, A. S. (2003). Music training improves verbal but not
visual memory: cross-sectional and longitudinal explorations in children. Neuropsychology, 17(3), 439-450.
Hotopf, M., Churchill, R., & Lewis, G. (1999). Pragmatic randomised controlled trials in psychiatry. The British journal of psychiatry, 175, 217-223.
Hu, L. t., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1-55.
Hulme, C., Bowyer-Crane, C., Carroll, J. M., Duff, F. J., & Snowling, M. J. (2012). The causal role of phoneme awareness and letter-sound knowledge in learning to read: combining intervention studies with mediation analyses. Psychological Science, 23(6), 572-577. doi: 10.1177/0956797611435921
Huss, M., Verney, J. P., Fosker, T., Mead, N., & Goswami, U. (2011). Music, rhythm,
rise time perception and developmental dyslexia: perception of musical meter predicts reading and phonology. Cortex, 47(6), 674-689. doi: 10.1016/j.cortex.2010.07.010
Jones, M. R., & Boltz, M. (1989). Dynamic attending and responses to time.
Psychological Review, 96(3), 459-491.
A r t i g o s | 116
Kolinsky, R., Lidji, P., Peretz, I., Besson, M., & Morais, J. (2009). Processing interactions between phonology and melody: vowels sing but consonants speak. Cognition, 112(1), 1-20. doi: 10.1016/j.cognition.2009.02.014
Lamb, S. J., & Gregory, A. H. (1993). The relationship between music and reading in
beginning readers. Educational Psychology, 13(1), 19-27. Lawley, D., & Maxwell, A. (1962). Factor analysis as a statistical method. Journal of the
Royal Statistical Society. Series D (The Statistician), 12(3), 209-229. Moreno, S., Bialystok, E., Barac, R., Schellenberg, E. G., Cepeda, N. J., & Chau, T.
(2011). Short-term music training enhances verbal intelligence and executive function. Psycholical Science, 22(11), 1425-1433. doi: 10.1177/0956797611416999
Moreno, S., Marques, C., Santos, A., Santos, M., Castro, S. L., & Besson, M. (2009).
Musical training influences linguistic abilities in 8-year-old children: more evidence for brain plasticity. Cerebral Cortex, 19(3), 712-723. doi: 10.1093/cercor/bhn120
Muthén, L., & Muthén, B. (1998-2012). Mplus (Version Seventh Edition ). Los Angeles,
CA: Muthén & Muthén . Nunes-Silva, M., & Haase, V. G. (2012). Montreal Battery of Evaluation of Amusia.
Validity evidence and norms for adolescents in Belo Horizonte, Minas Gerais, Brazil. Dementia and neuropsychology, 6(4).
Pasquali, L., Wechsler, S., & Bensusan, E. (2002). Raven's colored progressive matrices for children: a validation study for Brazil. Aval Psicol, 1(2), 95-110.
Patel, A. D. (2008). Science & music: talk of the tone. Nature, 453(7196), 726-727. doi: 10.1038/453726a
Pereira, L. (1981a). A coruja e a águia (Vol. Fourth grade). Sao Paulo: Editora do Brasil. Pereira, L. (1981b). O veterinário maluco (Vol. Third Grade). Sao Paulo: Editora do
Brasil. Pereira, L. (1993). Processamento Auditivo [Auditory Processing]. Temas em
Desenvolvimento, 2(11), 7-14. Peretz, I., Champod, A. S., & Hyde, K. (2003). Varieties of musical disorders. The
Montreal Battery of Evaluation of Amusia. Annals of the New York Academy of Sciences, 999, 58-75.
Peretz, I., & Coltheart, M. (2003). Modularity of music processing. Nature neuroscience, 6(7), 688-691. doi: 10.1038/nn1083
Peretz, I., & Zatorre, R. J. (2005). Brain organization for music processing. Annual
review of psychology, 56, 89-114. doi: 10.1146/annurev.psych.56.091103.070225
Slevc, L. R., & Miyake, A. (2006). Individual differences in second-language proficiency:
does musical ability matter? Psychoogical Science, 17(8), 675-681. doi: 10.1111/j.1467-9280.2006.01765.x
A r t i g o s | 117
Strait, D. L., Hornickel, J., & Kraus, N. (2011). Subcortical processing of speech regularities underlies reading and music aptitude in children. Behavioral and brain functions 7, 44. doi: 10.1186/1744-9081-7-44
Wechsler, D. (1991). WISC-III: Wechsler intelligence scale for children: Psychological
Corporation San Antonio, TX. Ziegler, J. C., Pech-Georgel, C., George, F., & Foxton, J. M. (2012). Global and local
pitch perception in children with developmental dyslexia. Brain Language, 120(3), 265-270.