Unlike in the US and Commonwealth jury systems, where judges give defined or undefined `beyond reasonable doubt' instructions, in England and Wales the juries are normally directed by the judge that the prosecution must make them sure of the defendant's guilt before they can convict. Critics of this paraphrase of the criminal standard of proof have adduced studies which appear to show that a high proportion of mock jurors equate sure with 100% confidence. Since it is impossible to be absolutely certain about evidence presented in court, the critics claim that jurors find themselves unable to convict. This paper critiques that argument from a number of angles, showing that the `sure as impossible standard' thesis is based on a misconstrual of the criminal standard of proof, on very weak empirical evidence, and on a misconceived notion of the relation between evidence and the linguistic expression of confidence. Drawing on a wide range of fields, as well as data from the author's extensive corpora of courtroom transcripts, the paper draws a distinction between formal and functional certainty and warns against the conventional wisdom which assumes commensurability between linguistic expression and numerical measurement.