Dave Liquorice wrote:

If the "normal" speech has a dynamic range of 10 to 20 db a
compressor set at 2:1 to 3:1 with around 6 dB of gain reduction
happening ought to work reasonably well.

Dave, since my previous reply I've sussed out the Audacity compressor (it
took me a shameful while to realise that the reduction was marked on the
display's horizontal axis rather than the vertical one). I've been trying
various settings and discovered that for the material I'm working with I
need rather more than 6 dB reduction: 20 dB appears to be nearer the mark.
That's if I want to retain a good contrast between the loudest and softest
passages.

I'll keep experimenting.