The soundness of the traditional Tarskian definition of logical consequence and, a fortiori, of logical validity, calls for a positive solution to the following problem. Given a language, is it possible to split it up into two mutually exclusive categories, that of logical expressions (topic-neutral) and that of non-logical ones (theory-laden)? In other words, is it possible to state an absolute criterion of logicality? Different formal approaches led to different definitions of what counts as a logical constant. At first glance, the choice among them could be considered as an issue concerning pure conventionalism. The aim of this paper is, on the one hand, to inquire into the legitimacy of such an idea and, on the other hand, to detect those situations in which speaking about pure logical conventionalism is completely justified.
