Through a series of judicial decisions and Practice Directions, the English courts have developed a rule that expert evidence must have ‘a sufficiently reliable scientific basis to be admitted’. There is a dearth of case-law as to what degree of reliability is ‘sufficient’. This article argues that the test should be interpreted as analogous to one developed in the law of hearsay: expert evidence (scientific or otherwise) must be ‘potentially safely reliable’ in the context of the evidence as a whole. The implications of this test will vary according to the relationship between the expert evidence and the other evidence in the case. The article identifies three main patterns into which this relationship falls. Whether the jury relies upon the evidence will depend upon what they regard as the best explanation of the evidence and how far they trust the expert. Whether their reliance is safe (as a basis for conviction) depends on whether they could rationally rule out explanations consistent with innocence, and whether the degree to which they take the expert’s evidence on trust is consistent with prosecution’s burden of proving the essential elements of its case, including the reliability of any scientific techniques on which it relies.