Abstract
An experiment has been carried out on the recognition of the orientation of ten different types of arrow for use in traffic signals. The main conclusions are as follows: 1. A conventional arrow designed at the RRL on the basis of previous research gives fewer errors than the design at present in use. 2. An unconventional design variants of which are used in the USA, Germany etc. is as good as the RRL arrow by night but less good by day. Modifications might improve its daylight performance but further experiments would be required to verify this.