Investigação psicométrica de uma tarefa de leitura em voz alta de palavras isoladas

• Ano de defesa: 2008
• Autor(a) principal: Patricia Silva Lucio
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: grafema-fonema e fonema-grafema; Fidedignidade; Reconhecimento de palavras; Validade; Análise de itens; Efeito de regularidade; Leitura; Psicometria; Psicologia do desenvolvimento; Psicologia
• Link de acesso: http://hdl.handle.net/1843/VCSA-7NYTS7

The reading word aloud task is generally used in Cognitive Psychology to evaluate word recognition and to test the components of dual-route models. The present study reports the psychometrical inquiry of a word reading aloud task composed by 323 low frequency words. A group of 1st to 4th grade school children from Belo Horizonte City (N = 333) read aloud the single words that were presented in a computer scream.Measures of reaction time (RT) and error scores were collected. The results indicated that in general, items were not difficult to the sample. The items that cause some difficulty to the sample (p<.40) were irregular ones. The Item analysis selected 112 potential items that presented satisfactory levels of discrimination and difficulty. There was a fatigue effect in RT, but was not demonstrated a training effect in error scores.The Generalized Estimating Equations (GEE) method exhibit the grapheme-phoneme and phoneme-grapheme regularity effect in reading and also showed an impact of the number of categories of regularity in this effect. A simulated item analysis used for data reduction showed that the pattern of the items obtained was similar to that found in theitem analysis, in terms of levels of regularity and size of the words. The item analysis simulation produced two versions of a task-example. The coefficients obtained by the method of alternate-form administrated in separated occasions indicated a satisfactory level of temporal stability for both tasks-examples. There was also reliability between the judges in conferring scores of those tasks. The results obtained for the itemdifficulty and discrimination analysis are discussed in relation to the differences between the two types of construct validity research proposed by Embretson (1983) the construct representation research and the nomothetic span research. We concluded that the psychometrical indicators produced in the present study contributed to constructvalidity under inquire.