This study conducts three experiments on learning layout descriptions and testing them by using direction labels versus arrow-keys. The pragmatics and requirements involved in each modality explain the accessibility patterns obtained with pointing and verbal labeling.