The redemption of truth: Idealization, acceptability and fallibilism in habermas' theory of meaning1

Abstract

Jürgen Habermas has proposed a tripartite classification of analytic philosophy of language into formal semantics, intentionalistic semantics, and use‐theories of meaning. Here, I focus on the relationship between formal semantics and Habermas’ own account of meaning and truth. I argue against his early ‘consensus theory of truth’, according to which truth is defined as idealized warranted assertibility and explained by the ‘discursive redemption’ of validity claims. A claim is discursively redeemed if it commands rationally motivated consensus of all discursive partners. I argue that this is not so much a theory of truth as of justification. As an account of truth, it is unsatisfactory because warrants can be lost in light of new arguments, and because consensus is seen as criterial for truth only if it is in principle defeasible ‐ hence there remains a gap between truth and even idealized warranted assertibility. In The Theory of Communicative Action, Habermas’ views undergo a shift. Following Dummett, he cashes out meaning in terms of assertibility conditions. While Habermas continues to insist on an epistemically conditioned notion of truth, he no longer explicitly defines truth as idealized warranted assertibility; nor does his meaning‐theoretic framework require him to do so. I argue that truth is an irreducible and ineliminable regulative idea. In particular, it can be explicated in terms of justification only indirectly. Contrary to some commentators, I do not believe that the regulative function of truth is merely negative. Fallibilism does not exhaust the meaning of truth. For we do not make truth claims provisionally, that is, with the understanding that they may turn out to be wrong. Rather, we gamble that we will not turn out to be wrong. Therefore, the concept of truth has not only a critical, but also a positive function in our discourse.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 214 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.