We consider a novel null test for contamination which can be applied to CMB polarization data that involves analysis of the statistics of the polarization position angles. Specifically, we will concentrate on using histograms of the measured position angles to illustrate the idea. Such a test has been used to identify systematics in the NVSS point source catalogue with an amplitude well below the noise level. We explore the statistical properties of polarization angles in CMB maps. If the polarization angle is not correlated between pixels, then the errors follow a simple $\sqrt{N_{pix}}$ law. However this is typically not the case for CMB maps since these have correlations which result in an increase in the variance since the effective number of independent pixels is reduced. Then we illustrate how certain classes of systematic errors can result in very obvious patterns in these histograms, and thus that these errors could possibly be identified using this method. We discuss how this idea might be applied in a realistic context, and make a preliminary analysis of the WMAP7 data, finding evidence of a systematic error in the Q and W band data, consistent with a constant offset in Q and U.