Still, language has its pitfalls as a guide to national lines of cleavage. To begin with, while in the country districts it is singularly persistent, it can also be learned and unlearned by a new generation, especially when the resources of universal education are wielded by the compulsive power of the modern state. The German schoolmaster has labored to reduce the area of Danish in Schleswig and of French in Alsace-Lorraine. The Russians have checked German in the Baltic provinces. In the British Isles Celtic speech is far less widespread than Celtic blood. Moreover, if the government cannot always drive out minority languages, it can at least make its own language statistics. It is easy for the census-taker to impose on the weak or ignorant, to interpret all doubtful cases in one direction, to adopt definitions which fall on one side. The statistics of Polish-speaking districts in Prussia and of Italian-speaking elements in Austria-Hungary are well known examples. Again, language statistics, even when trustworthy, do not necessarily yield a sharp dividing line. There may be more than two significant elements in the population, or, as in parts of the Balkans and Asia Minor, villages of different speech may be interspersed in a checkerboard fashion. Finally, language, even when accurately ascertained, is not a certain test of political affiliation. There has been a strong pro-French tradition among the German-speaking Alsatians. The small German-speaking districts of Belgium are not pro-German. After all, it is the opinion of the people concerned which we wish to ascertain with respect to a given frontier, and language is important chiefly as a guide to that. Unequivocal expressions of popular opinion are, however, hard to reach, especially in times of stress and in regions that are under dispute. At best a plebiscite may be but a poor indication of real opinion, and the opinion it registers well may be only transitory. Moreover, caution may be required in giving effect to a vote or an otherwise well ascertained expression of opinion. Thus it is open to debate whether the vote of a small district should necessarily carry with it the disposal of a great key deposit of mineral wealth which concerns a much wider constituency. It is also possible that in the long run commercial intercourse and economic interest may create ties more lasting than language or national sentiment, and that a given boundary may do more ultimate harm by violating the fundamental economic interests of a region than by violating its momentary political sentiments.