Dawkins lays into the phrase, and suggests an experiment to show what a sham it is:
if [a] two-jury experiment were run over a large number of trials, the frequency with which two groups would agree on a verdict would run at slightly higher than 50 per cent. But anything short of 100 per cent makes one wonder at the “beyond reasonable doubt” held to be sufficient to send somebody to the electric chair.
we should acknowledge that “beyond reasonable doubt” is a hollow and empty phrase. If you defend the single-jury system as delivering a verdict “beyond reasonable doubt”, you are committed to the strong view, whether you like it or not, that two juries would always produce the same verdict. And when you put it like that, will anybody stand up and bet on 100 per cent concordance?