Spatial knowledge was evaluated in sighted and congenitally blind children using a large-scale four-location navigation task adapted from the work of Landau, Spelke, and Gleitman (1984). From video records we coded the exact path taken and determined accuracy of initial turn, closest position, and final position, relative to target location. We then computed a score to index the efficiency of the path taken. For the sighted sample, after the navigation task, children constructed a tactile map of the test space without the aid of vision and, following removal of the blindfold, drew from memory the spatial layout of the test space. Performance on the navigation and mapping tasks consistently indicated increasing cognitive mapping skills with age in sighted children. Blind children performed comparably to the sighted on all measures except accuracy at final position, for which their performance was worse than that of the sighted. Analysis of the directness of novel paths and other measures taken suggest caution in ascribing well developed Euclidean coding skills to very young children. Results are discussed in light of Landau et al.'s (1984) conclusions.