The Hotelling T 2 control chart is widely used to monitor the process mean vector in multivariate quality control. However, this control chart is relatively insensitive to small and moderate shifts in the process mean as it uses information only from the current sample. A new type of control chart is proposed to detect shift of the mean of a vector of two quality characteristics which are normally distributed. The acceptance region for declaring that the process is in control is a circle of which the centre depends on the mean based on the all the previous data. By using simulation, the performance of the new chart as measured by the average run length is found to be better than that of the Hotelling T 2 control chart.