Pre-modern socities which possess agriculture and literacy, the inhabitants of what Gellner sometimes calls "Agraria," were economically static and internally culturally diverse, at least compared to their industrial successors. Cultural differences in fact often went with economic specializations, and so served to fix people in their inherited professions. It is Gellner's thesis that economic change requires cultural homogeneity, and that the demand for cultural homogeneity, and the state apparatus to provide it, is what drives nationalism. The argument runs as follows.
Because industrial economies continually make and put into practice technical and organizational innovations, they continually change how they employ resources, especially human resources. Their occupational structures change significantly in a generation at most, and often more quickly, so no one can expect to follow in the family profession. (A hundred years ago, there were no system administrators, but there were carriage-drivers.) In Agraria, training could be left to families or guilds, be largely tacit and physical and tied up with the rituals and social context of the trade, and different parts of the same society could be almost unintelligible to each other, provided only they could go through the customary haggling or tithing. None of this will do in an industrial, changing society, in which training must be much more explicit, be couched in a far more universal idiom, and emphasize understanding and manipulating nearly context-free symbols (even manual work increasingly becomes controlling a machine, which must be, as we say, read); it must in short take on the characteristics formerly associated with the literate High Cultures of Agraria, and moreover this training must be received by the entire economically effective population. (A rough definition of an industrial society might be: one where you can learn a trade from books, a society of reference manuals.) So far, such training, on such a scale, has always needed at least elementary literacy, and it hasn't been reliably provided by any institutions weaker and smaller than states. Moreover, the teachers employed by this system must themselves be trained in the same High Culture, and so on, quickly escalating to the point where the culture needs an entire university system, at the least, to be self-sustaining. States become the protectors of High Cultures, of "idioms"; nationalism is the demand that each state succor and contain one and only one nation, one idiom.
To be without such an idiom is to be cut off from all prospects of a decent life. To have the wrong idiom, that is, a different one than those in charge about you, adds the constant humiliations of being a stranger, outcast, isolated, constantly doing "the wrong thing" --- quite possibly while knowing that one's own ways would work at least as well. Thus the passion behind nationalism derives, not from some atavistic feeling of tribal belonging (supposing such a thing to exist at all, outside of the immediate circumstances of mass rallies and the like), but from the hope of a tolerable life, or the fear of an intolerable one. Faced with an difference between one's own idiom and that needed for success, people either acquire the latter, or see that their children do (assimilation); force their own idiom into prominence (successful nationalism); or fester. Thus industrialism begets nationalism, and nationalism begets nations.
This last point requires more emphasis. Agraria was a mess of partially overlapping ethnic, religious, linguistic, political and cultural distinctions. On the issue of language alone, Gellner calculates that the Old World contained several thousand dialects, each of which could have been the basis of a formalized literary language. (This calculation excludes Papua New Guinea.) Nations are constructed, in a highly arbitrary way, out of this raw material, often with a great deal of false consciousness (e.g., thinking one is reviving peasant culture and folk traditions, while actually creating a formalized, school-dependent High Culture) and outright fabrication. It is an error to suppose that nations have always existed, or even that modern nations are very old. (Serbo-Croat, for instance, was created as a literary language in the 19th century; in the last decade it has "officially" become two languages, Serb and Croat, which obstinately persist in being mutually intelligible.)
To recap: industrialism demands a homogeneous High Culture; a homogeneous High Culture demands an educational system; an educational system demands a state which protects it; and the demand for such a state is nationalism. The theory is coherent, simple, widely applicable, convincing, and empirically testable (which tests, to all appearances, it passes).
In the sixties, it is said, the Right and the Left in Britain agreed on only one thing: Gellner would be up against the wall when the Revolution came. He was quite proud of that, and the same spirit comes through abundantly in this book. It is hard to decide whether nationalists or anti-nationalists will find Nations and Nationalism more disturbing; rootless cosmopolitan though I am, it changed my mind on a great many subjects. This is already a rare enough achievement for a philosopher or social scientist, let alone someone like Gellner, who was both, since they typically change no one's mind. Even if Gellner was talking utter rubbish, Nations and Nationalism would be worth reading simply for his style, a trademark Bertrand-Russell-meets-Grucho-Marx combination of powerful logic working from very general premises and laugh-out-loud (literally) wit. Unfortunately for those of us not enamoured of nationalism, he wasn't talking rubbish at all.