These developments reflect the problem of how to decide what is an error.
Step 3 of the methodology described earlier solves this by treating the input
copy as if it were fully accurate, and identifying any non-congruence in
output copy as bias or inaccuracy. We thus limit our field to the smaller
compass of copy editing stages (rather than a final report versus "the facts"),
but by that means sharpen our focus and strengthen our inferences.
A second strand of research emerges in recent work in Britain and
Australia. Here linguistic analysis is used to draw inferences concerning the
ideological biases of certain media. The most publicized research (e.g. The
Listener, 29 July 1982) is that of the Glasgow University Media Group (1976,
1980). The Group periodically gets into debate with BBC television over
bias in the BBC's industrial news, or alternatively, in the Glasgow Group's
views. The Group has noted (1976, p. 256) inaccurate editing changes
between successive news bulletins which are of precisely the kind present in
my data.