We present a simple test for normality against the alternative that the underlying distribution is long tailed. The test is based on the behaviour of the empirical characteristic function in the neighbourhood of the origin, and is very competitive with several well-known tests for normality.