College-student subjects engaged in judgments of numerosity received stimulus arrays on a computer screen that varied in their number level (low, medium, and high), the speed with which the array developed (slow, medium, or fast), and the density of the items in the array (low or high density). The subjects overestimated number when the array was small (75-125 items) but significantly underestimated number when it was at a medium (150-250) or high level (550-650). Density of array items was a significant variable, since subjects showed more underestimation under the high-density condition. Additionally, there was a speed x density interaction. The results are discussed with respect to which class of variable, sensory or cognitive, most influences perceived numerosity. Although they deal with the animal literature, Davis and Perusse (1988), in their excellent review, distinguish among a number of approaches to the measurement of numerical competence. One of the approaches, estimation, is different from subititing because the subject is re