Text comprehension in second and third graders: a cognitive approach. This study aimed to analyze text comprehension of students of the 2 nd and 3 rd grades. The sample was constituted by 76 children, at an average of 8.1 years old. Each child read the story, retold it and, afterwards, answered questions about it. The retellings were analyzed according to the model of Text Comprehension of Kintsch and van Dijk (1978) and Kintsch (1988Kintsch ( , 1998. The sample recalled a mean of 21.07% of the proposition structure of the story, being the report of macropropositions more frequent. Students of the third grade told larger percentage of irrelevant micropropositions of the text and they were superior in answering to specific questions than students of the second grade. A significant correlation was found between age and macroproposition's retelling. The results suggest that during the first years of schooling there is an improvement of the detail-remembering, whereas the retention of the essential ideas is influenced by age differences.Key-words: text comprehension; reading; story retelling; reading evaluation Cantor, & Carullo, 1992;Swanson & Alexander, 1997;Waters & Caplan, 1996), exposição à escrita, conhecimento do vocabulário, metacognição (Cain, 1999;McBridge, Manis, Seidenberg, Custodio, & Doi, 1993), habilidades semânticas (Nation & Snowling, 1998), consciência sintática (Gaux & Gombert, 1999), além de processos que também são usados para a compreensão da linguagem falada, incluindo o conhecimento geral sobre o tema tratado no texto (Ellis, 1995). Braibant (1997) propõe três fatores essenciais para explicar a presença de dificuldades de compreensão, quando as capacidades de identificação de palavras são eficientes: (1) capacidades lingüísticas (vocabulário, sintaxe) e cognitivas (elaboração de inferências) insuficientes; (2) falta de automatização na captação de informação escrita e (3) incapacidade para adaptar suas estratégias ao objetivo procura-