Some authors consider classes as communities (Schumpeter, Sorokin) or groupings of families (Erikson 1984). Others consider them as conditions (Dahrendorf 1959; Runciman 1969), positions, or roles (Gallino 1987) assumed in society. Such theorical uncertainty is followed by a similarly uncertain empirical classification: Poulantzas (1973) says that society is divided in three classes; Roemer (1982) mentions five classes; Wright (1985, 120) draws a typology of twelve positions of class.
Why this confusion? Probably because classes aren't, in first instance, ostensible objects but concepts, i.e. culturally-and-mutually-constructed cognitive schemas. In order to see classes scientists have to agree about the culturally framed discourse to use. But at the moment this hasn't happened yet. This seems the main cause of the endless conflict in the debate on social stratification.
Further, the essay documents as `class', before being scientific construct, was a 'folk category'. From ordinary language 'class' reaches the social sciences, passing through the natural sciences. Scientists would have done anything more than specialize non-scientific linguistic uses. In this way common-sense prejudices and stereotypes, which distinguished the concepts of `class' in ordinary language, would have filtered into the sciences.