The number of speakers of a language can be of importance for educational planning, as economic factors may forbid the inclusion of too small a language in the school curriculum, and speakers of this language would then be forced to use a second language. However, establishing the true number of speakers of a language can be problematical as (1) the concept "speaking a language" would need an exact definition, and (2) a method would have to be evolved for eliciting accurate information corresponding to the work- ing definition. Censuses are not always reliable. In Ireland and Wales, for example, they do not differentiate between those who have Irish or Welsh as their mother tongue and those who have acquired the language later in life, e.g. at school or in evening classes {Price, 1973). Similarly, the Lapps living in the coastal districts of Northern Norway refused to admit their true identity in a census, as Lapp is considered a negative identity in Norway (Aarseth, 1969). However, even a linguist conducting a personal interview is not free from error, as the interviewees will often give the answer expected or hoped from them rather than the true answer. In the case of speakers of a language with low prestige there will be a tendency to negate one's language competence, and in the case of a minority language where the interviewer is obviously a supporter of the language, there will be a tendency to overestimate one's linguistic ability. One possible method of checking the true linguistic competence of an interviewee is to converse in the given language. If the interviewer switches codes, however, it is possible that the interviewee will not be able to follow suit if he has the linguistic rule 'one person one language'. {I have experienced this in my field work in NF.)