It is easy to think, without having sufficient evidence to support the opinion, that a type of behaviour shows a propensity. For example, suppose that D is charged with burglary. A witness could only say that the burglar carried a red bag. The defendant owns a red bag and on two previous occasions had used it while committing burglaries. This looks like a propensity. But what if further evidence showed that D had committed eight other burglaries while carrying a black bag. Is D’s propensity to burgle with a black bag or with a red bag? And is either proved beyond reasonable doubt to be a propensity?
Furthermore, using the same example, evidence may be more probative “of itself” than if it is used as propensity. Forget the other burglaries and ask how many of the possible present burglars have red bags? This might be a very small proportion. As the denominator it would make the Bayesian likelihood ratio large enough to give D’s possession of a red bag significant probative value. Perhaps this demonstrates that propensity, if treated in the Mitchell way, can be something of a red herring, detracting from the real value of the evidence that is directly relevant to the present allegation.