OBJECTIVE To investigate the capacity for self-monitoring reading comprehension in Brazilian Elementary School students. METHODS Fifty-three Elementary students in the 5th and 9th grades from two Public Schools in the city of São Paulo were assessed. They were selected based on their oral reading rate and grouped according to their performance in reading comprehension in the following categories: Group with best comprehension: students with adequate rate and accuracy, without difficulties in reading comprehension; Group with worst comprehension: students with adequate rate and accuracy but with difficulties in reading comprehension. Two narrative texts followed by eight questions to assess reading comprehension were presented. Two sentences and two words were replaced by ungrammatical elements and pseudo-words. Under the condition of spontaneous monitoring, students read the text aloud and answered the questions. The analysis considered the calculation of hesitation, self-correction, repetitions and mistakes. Under the condition of directed monitoring, students were instructed to read the text, either aloud or silently, after being told that certain parts of the text could not make sense, and they were oriented to underline such parts. The analysis was carried out by counting of underlined items. The comparisons were made with the Mann-Whitney test. RESULTS A difference was observed between the groups only at the sentence level among the 9th grade schoolchildren under the spontaneous monitoring and among the 5th grade schoolchildren under directed monitoring. CONCLUSIONS Students with worst comprehension had a poorer performance to monitor the presence of ungrammatical sentences than their peers with best comprehension.