<p>Some types of forensic evidence are analysed by forensic experts using non-probabilistic methods; the conclusions of the analysis are then reported to courts as categorical opinions. Fingerprint evidence is the most common example. The article calls for a change of practice: forensic evidence should be analysed using probabilistic methods and reported to the court in non-categorical terms. It should then be for the court to decide which categorical conclusions to draw from the evidence. The court should do so based on the unique practical considerations that enter at that stage and that guide legal decision-making. It is suggested that forensic scientists are not well-placed and indeed should not be authorised to weigh such considerations, which are inconsistent with those relevant for scientific discourse.</p>