Roughly stated, conceptions of language underlying the analysis and interpretation of 'things' are divisible into two categories: language is either considered inadequate as a source of knowledge or, on the contrary, it is seen as the only source of knowledge we have, and a sufficient one. In Plato's dialogue Cratylus, the first point of view is foreshadowed by Cratylus, who holds that only names that are naturally right are real names. This turns most language into 'mere noise'. Cratylus's opponent Hermogenes is prototypical of the second point of view. He regards all names as based on conventional agreement and therefore as equally real. Language always makes sense. This divergence of views on the boundaries of language is reduced to a debate on knowledge by a third interlocutor, Socrates. Socrates conceives both contradictory conceptions of language as attempts to solve one problem, which rises because of his presupposition that language consists of true and untrue judgments. For Socrates, distinguishing between truth and untruth is a problem of knowledge about 'things', not a linguistic one. When it comes to making the distinction, he ironically admits his ignorance. This may be wisdom, but it doesn't solve the problem of distinguishing language from nonlanguage.
   Through the ages, this debate has been repeated in various ways. In the eighteenth century, Kant distinguished between analytic and synthetic judgments. He did not go into their characteristics as linguistic judgments, but restricted himself to their relationship to truth. Analytic judgments are logically true or untrue; their truth depends on their linguistic form. Synthetic judgments are based upon interpretation; their truth depends on empirical evidence. Cratylus's idea of an ideal language transcending the confusions of everyday life is thus linked up with Hermogenes's idea of a language recognizing the importance of everyday certainties. For the sake of truth, however, Kant needs judgments that are both logical and empirical, i.e. ideal and experiential at the same time – expressing a priori knowledge of synthetic truths. Like Socrates, Kant has recourse to knowledge of 'things', Dinge an sich, in order to avoid questions of language. Like Socrates, he presupposes the existence of a pure reason excluding differences of opinion. Against practical efforts to demarcate language in different ways, Socrates and Kant put their ideal of a pure knowledge that cannot be argued against.

The socratic-kantian ideal of knowledge is at the basis of virtually every modern theory of language. The 'analytic' point of view of Cratylus and his descendents reacts to the idea that everyday language is imperfect by trying to compensate for the defects. It attempts to reach a better organization of language: to come closer to 'things' (like some mystics, poets, madmen, and scientists), to be formally more consistent (e.g. by means of arithmetics, logic, computer languages), or to communicate more purely (e.g. in the standard language of Esperanto). Such struggles for organization are manifest in Bernstein's sociolinguistics, Heidegger's language that is 'open to Being', Frege's formalization leading to problems with judgments of identity, and Tarski's even more consequent formalization demonstrating that looking for truth doesn't lead to very interesting judgments at all. From a comparison of these authors (Ch. 2), one can conclude that, by trying to compensate for supposed defects of language, one also looses the 'metaphysical' values and certainties that speakers of a language used to share.
   The 'synthetic' point of view of Hermogenes and his offspring reacts to the idea that everyday language is imperfect by simply denying it. It attempts to show that the apparent disorganization of everyday language hides an organization that can be discovered through interpretation. For a willing ear, there is much to learn from authentic formulations of truth (by some spiritists, converts, and bureaucrats), holistic systematizations of language (e.g. by structuralists or cognitivists), and all-embracing forms of communication (by moralists, preachers, and conflict avoidance strategists). Such a denial of disorganization can be observed in the biblical 'speaking in tongues', which affords a glance at the language before the Babylonian confusion of tongues, and also in Labov's chomskyan reaction to Bernstein, in Grice's principle of cooperation, and Habermas's idealized communication. A comparison of these authors (Ch. 3) shows that a reintroduction of shared knowledge and understanding tends to mis-appropriate language that differs from one's own or to deny the existence of such deviant language.

The 'analytic' and 'synthetic' approaches of language are complementary. If one tries to speak in a way that is universally true, then, in the end, all traditional meaning gets lost. If, on the other hand, one tries to find meaning somewhere or everywhere, it cannot be deemed undeniably universally true. In itself, this complementarity is not problematic. It becomes problematic in the light of the socratic-kantian presupposition that one should be able to distinguish in a pure way between truth and untruth. Language is no means to do that. The reduction of theories of language to theories of knowledge, therefore, leads to confusions and vicious circles.
   In order to change this unsatisfactory state of affairs, I propose to stop looking for compromises between both solutions, and to contribute to a radical alternative: a theory of language that is skeptical about claims of knowledge – in the sense of Pyrrho and Sextus Empiricus. As opposed to Socrates's ironical doubt about the accessibility of truth, which leaves the norm of pure knowledge as it is, pyrrhonian skepticism denies that knowledge matters. Thus, it makes room for a return from theories of knowledge to a theory of language. A skeptical theory of language can engage in language without pretending to know more than it does: it need not ascribe knowledge (understanding, communication, etc.) to speakers and/or hearers, but it can – continuously – add to their conversations by describing what happens in linguistic exchanges (under what circumstances they gave a feeling of acquaintance, of solidarity, or of drifting apart). Involvement might be a more fruitful contribution than knowing it all.