This article describes an experiment designed to explore the effects of 'priming' (i.e. being exposed to a suggested interpretation of an audio signal) on how juries perceive disputed utterances in poor quality recordings used as evidence in legal cases. Using the actual disputed utterance from a real case, the experiment tracks how participants' perception of its content changes as evidence about the case is gradually revealed to them. At a certain point, participants are randomly divided into two groups, each receiving parallel but slightly different evidence. Results indicate the dangers of priming may be considerably greater than is sometimes recognised, and unlikely to be overcome by a mere caution from the judge. They also indicate that participants' propensity to consider the defendant guilty may be based on judgements of his trustworthiness influenced by initial impressions of his style of speech, rather than on objective evidence presented to them.