This paper examines to what extent police investigators can reliably question a vulnerable suspect’s account when the evidence base for appropriate questioning styles for this particular vulnerable group is limited. We examine a simulated interview to demonstrate how difficult it is to challenge discrepancies in a vulnerable suspect’s account. It is argued from both linguistic and psychological perspectives that certain question formats may lead to acquiescence, cognitive overload and confusion for the suspect. It is suggested that one way of trying to manage these issues is through the provision of alternative narratives (i.e. ‘only one of those stories can be true’) but these too are found to be problematic.