Two groups of twelve dyslexic children and twelve good readers attended a two-week class during the summer following their enrollment in grades 4-6. The topic was the marine environment and focused on the adventures of an orca whale that had lost its family and needed the children's help. The children attended class for three hours each day, 45 minutes of which were devoted to working with a computer-based, interactive, three-dimensional simulation of the ocean. The rest of the time was spent on hands-on, problem-based, non-verbal classroom activities. Data obtained from logs of the children's activity using the computer, maps drawn on two occasions from memory, and brain activation obtained from fMRI before and after the class showed that the dyslexic children were as able to use the simulation and construct spatial mental models as the good readers. However, significant correlations among activation of areas of the brain associated with working memory were found during fMRI imaging for good readers performing a 2-back visual memory task, but not for dyslexics. Significant correlations were also found among time taken for dyslexics, but not good readers, to complete computer-based exercises and activation of area V5 while they performed tasks requiring visual decoding of pseudowords. These findings, which suggest that good readers and dyslexics might use their brains differently to reach similar levels of performance, were interpreted within cognitive and neurocognitive theories of learning.