Physically edited copy was available for many stories analysed from my sample (Figure 2 is a page of wire copy with a recipient editor's markings.) For most of the other stories used, the input was identified through internal evidence. These were once-off stories, or reports containing idiosyncratic phrasing or information (such as figures). In other cases, it is impossible to decide whether an output story is in fact derived from a certain input wire. Big stories may have 10-20 continually updating wires, with a broadcast news item which is a rewrite combining several wires. The form of a few input wires was reliably reconstructed by comparing several output versions. In all cases, the necessary principle followed was: when in doubt, omit. When the output is broadcast news, we use the form actually read out by the newscaster and tape-recorded, which often differs from the edited script. The eventual data for this study included 290 international stories, with both input and ouput copy.