Increasingly, computers are becoming tools of communication, information exploring and studying for young people, regardless of their abilities. Scientists have been building knowledge on how blind people can substitute hearing or touch for sight or how the combination of senses, i.e., multimodalities, can provide the user with an effective way of exploiting the power of computers. Evaluation of such multimodal user interfaces in the right context, i.e., appropriate users, tasks, tools and environment, is essential to give designers accurate feedback on blind users' needs. This paper presents a study on how young blind people use computers for everyday tasks with the aids of assistive technologies, aiming to understand what hindrances they encounter when interacting with a computer using individual senses, and what supports them. A common assistive technology is a screen reader, producing output to a speech synthesizer or a Braille display. Those two modes are often used together, but the research studied how visually impaired students interact with computers using either form, i.e., a speech synthesizer or a Braille display. A usability test has been performed to assess blind grade-school students' ability to carry out common tasks with the help of a computer, including solving mathematical problems, navigating the web, communicating with e-mail and using word processing. During the usability tests, students were allowed to use either auditory mode or tactile mode. Although blind users most commonly use a speech synthesizer (audio), the results indicate that this was not always the most suitable modality. While the effectiveness of the Braille display (tactile user interface) to accomplish certain tasks was similar to that of the audio user interface, the users' satisfaction rate was higher. The contribution of this work lies in answering two research questions by analysing two modes of interaction (tactile and speech), while carrying out tasks of varying genre, i.e., web searching, collaboration through e-mail, word processing and mathematics. A second contribution of this work is the classification of observations into four categories: usability and accessibility, software fault, cognitive mechanism and learning method. Observations, practical recommendations and open research problems are then presented and discussed. This provides a framework for similar studies in the future. A third contribution of this work is the elaboration of practical recommendations for user interface designers and a research agenda for scientists.