Who are Roma? What is the size of Roma population in a given country, a region of the world or in the whole word? To date, most social scientific research done in the Central European region on Roma populations operated with the assumption that one can arrive at an objective definition who Roma are and thus to come uup with accurate estimate of the size of Roma population. Rather than aiming at an „accurate” estimation of the Roma population, in this paper we hypothesize that ethnic groups are „social constructions”. The boundary of any single ethnicity is „fuzzy”: who is „inside” and who is „outside” this boundary will vary depending who does tha classification. This paper tests survey results on two systems of classification: 1) self-identification by the repondents, and 2) and classificaton by interviews in three countries, Bulgaria, Hungary and Romania; and we show that there are substantial variation across countries. The paper offers strong support to the hypothesis that ethnic boundaries and various systems of classification are alterble across cultures. About two-third of those two were classified as Roma by the interviewers in Hungary and Romania do not regard themselves as Roma. In Bulagria two-third of the respodent who were classified as Roma by the interviewer identifier themselves as Roma. Therefore the task of social research is not to identify which classificatory system is „correct”, or „accure”, but to understand the social processes how these various classificatory system are created.