The main purpose of the present study was to investigate whether in natural environment, using very large physical distances, there is a trend to overconstancy for distance estimates during development. One hundred and twenty-nine children aged 5 to 13 years old and twenty-one adults (in a control group), participated as observers. The observer's task was to bisect egocentric distances, ranging from 1.0 to 296.0 m, presented in a large open field. The analyses focused on two parameters, constant errors and variable errors, such as measuring accuracy and precision, respectively. A third analysis focused on the developmental pattern of shifts in constancy as a function of age and range of distances. Constant error analysis showed that there are two relevant parameters for accuracy, age, and range of distances. For short distances, there are three developmental stages: 5-7 years, when children have unstable responses, 7-11, underconstancy, and 13 to adulthood, when accuracy is reached. For large distances, there is a two-stage development: 5-11 years, with severe underconstancy, and beyond this age, with mild underconstancy. Variable errors analyses indicate that precision is noted for 7 year-old children, independently of the range of distances. The constancy analyses indicated that there is a shift from constancy (or slightly overconstancy) to underconstancy as a function of physical distance for all age groups. The age difference is noted in the magnitude of underconstancy that occurs in larger distances, where adults presented lower levels of underconstancy than children. The present data were interpreted as due to a developmental change in cognitive processing rather than to changes in visual space perception.