> > I get the idea that it's the same for the so-called "r-sound", basically, a > symbol for any rhotic when the exact quality is not known. Naturally, such > symbols are going to be a pain to translate into IPA, but could be useful > nonetheless. What should we do with them? > -- > Christophe Grandsire-Koevoets. > > http://christophoronomicon.blogspot.com/ > http://www.christophoronomicon.nl/ > How about this: choose the otherwise most frequent representation in the segment class, with a mark denoting this is an approximation, until the ensuing comments reach a conclusion? I know that I would prefer complete statistics over 100%-accurate but lower-sample ones.