Children ages 2, 3 and 4 years participated in a novel hide-and-seek search task presented on a touchscreen monitor. On beacon trials, the target hiding place could be located using a beacon cue, but on landmark trials, searching required the use of a nearby landmark cue. In Experiment 1, 2-year-olds performed less accurately than older children on landmark trials but performed equivalently on beacon trials. In Experiment 2, the number of items on the screen was reduced and 2-year-olds' performance improved. Use of the landmark transformation technique in Experiment 3 revealed that older children formed a more precise landmark-target spatial relationship than 2-year-olds. Experiment 4 showed that the transformation itself was not responsible for the youngest participants' decreased accuracy in Experiment 3. Overall, beacons were utilized effectively by all participants, but the use of landmark cues is refined between the ages of 2 and 4.