In this paper, I aim to raise a number of difficulties in Laurence Bonjour's coherence theory of epistemic justification and in doing so, I intend to provide special focus on difficulties that arise from the limitations of language as it is used in the discursive manner by which a philosophical theory is constructed. To make an initial clarification, what I have in mind is an image of a philosopher who first intends to create a theory and then proceeds to construct and strengthen his or her theory in a discursive and elaborative manner in which language inevitably plays an indispensable role. In conjuring this image, I am reminded of John Rawls' constructivist approach towards a theory of freedom and of Jürgen Habermas' discourse theory, both of which recognizes the paramount importance of language and discourse in the vitality of a philosophical theory as well as in the methodological manner by which theorizing and philosophizing are carried out.
Now I also imagine Bonjour alone in his office, thinking first about foundationalist theories of epistemic justification and recognizing and elaborating on the weaknesses of foundationalism. Afterwards, he then attempts to characterize an alternative theory of epistemic justification, which is his coherence theory, and in doing so he makes use of language to present and develop arguments and elaborations in a discursive and analytic manner, all towards the objective of constructing a strong and robust coherence theory. He starts with the general idea of coherentism – the idea that a belief is linked to other beliefs in a coherent system and the justification of this belief can only be attributed to the coherent system as a whole and not to some particular basic or foundational beliefs. However, this initial general characterization is insufficient and so Bonjour must now proceed with a more elaborate characterization of his theory and address the difficulties that arise along the way.