On Tue, 18 Aug 2009, Allin Cottrell wrote: I've now taken a look at this issue (though having a sample datafile might still be helpful). The difficulty with handling non-ASCII labels in stata dta files is that character encoding is not handled properly in those files. The online stata documentation for the dta format says "Strings use ASCII encoding", which is a lie! Apparently, they use whatever encoding is in force on the platform where the dta file was created, but the file contains no record of the encoding used. See http://cluelessresearch.com/2007/05/stata-and-accents-diacriticals/ What gretl does with such labels is: * If a label validates as UTF-8 (which includes ASCII as a subset), we use the label as is. (I've verified that the Spanish letter 'enie' comes out right in dta labels if the dta file is created on a platform that uses UTF-8.) * Otherwise we try recoding the label from the current locale to UTF-8. This will work only if the current locale happens to be the same as that under which the dta file was created. * New in CVS: if the above does not work we guess that the encoding of the label might be Windows CP1252, and try converting from that. If none of the above work, you're out of luck. Complain to stata! Allin Cottrell