This experiment tests the age at which left-to-right spatial associations found in infancy shift to culture-specific spatial biases in later childhood, for both numerical and non-numerical information. Children ages 1 to 5 years (N=320) were tested within an eye-tracking paradigm which required passive viewing of a video portraying a spatial transposition. In this video, an item was hidden in a vertical set of locations, which were then surreptitiously rotated 90°. There were several conditions, which varied in the degree to which the locations were presented alongside ordinal (numerical, alphabetical) or non-ordinal (nonsense label) information. After transposition, a narrator prompted the child to visually search the array. The amount of time spent fixating in a location consistent with a left-to-right mapping or a right-to-left mapping was measured to gauge the degree and laterality of spatial associations. Overall, children looked more towards locations consistent with a left-to-right mapping. This effect fluctuated with age, dipping as children entered toddlerhood, increasing in 3- and 4-year-olds, and then disappearing at age 5. The ordinal nature of the stimuli (e.g., numerical or non-numerical) did not influence the laterality of the spatial associations. A follow-up experiment confirms that, like older preschoolers, adults (N=66) also exhibit no spontaneous left-to-right mapping bias in this paradigm, with no fluctuation as a result of condition. These data support the presence of a decrease in left-to-right processing around the age of two, as children recede from infantile spatial biases and progress to exhibiting culture-specific spatial biases in early childhood.