There’s now a fourth edition of this. I read the second edition, published just after Gutenberg got his printing press together, so these comments have to be considered in that light.
In all, this was a pretty dry read but had some worthwhile stuff which I gleaned from it which was of direct relevance to doing Language Survey. That said, I had to work at it in places as Fowler focusses mostly on census and marketing surveys in the USA. Sections on procedures for telephone surveys, for example, can be helpfully skipped over by the linguist in PNG!
The most relevant sections, I found, were those on sampling and question design. The former, he goes into in some detail and this was very helpful for me as I’ve absolutely no background in statistics at all. Not sure that I feel confident after reading this, but I’m certainly more aware.
to assume perfect correspondence between the answers people give and some other reality is naive
rubbish | poor | mediocre | okay | good | very good | excellent | superb