The current research examined whether peripherally presented numerical information can affect the speed of number processing. In 2 experiments, participants were presented with a target matrix flanked by a distractor matrix and were asked to perform a comparative judgment (i.e., decide whether the target was larger or smaller than the reference 5). In Experiment 1, the target was symbolic (i.e., a single digit), and in Experiment 2, it was nonsymbolic (i.e., a random presentation of dots). In both experiments, flanker matrices had 2 dimensions--numerosity and numerical value--that were manipulated orthogonally to create stimulus congruent and stimulus incongruent conditions. Incongruent trials differed in the laterality between target and flanker (i.e., their location in relation to the reference 5). When responding to symbolic targets (Experiment 1), only the flanker's numerical value affected reaction times (RTs), whereas when responding to nonsymbolic targets (Experiment 2), only the flanker's numerosity affected RTs. In addition, the pattern of flanker interference differed between targets: For symbolic targets, laterality did not affect responses, whereas for nonsymbolic targets, laterality did affect responses. These results imply both symbolic and nonsymbolic magnitudes can be automatically activated; however, this activation is contingent upon their relevance to the task at hand. Implications of these results on the efficiency of the visual processing system and on numerical cognition are further discussed.