BonJour summarizes his account of the fully explicit justification of a particular empirical belief on p. 228 of "Holistic Coherentism". There are four parts to the account. I have elaborated on them below, in some cases incorporating ideas from his later book (The Coherence Theory of Empirical Knowledge) and in some cases making my own attempts to explain what an item involves. On BonJour's account, it is necessary to distinguish the local justification of an empirical proposition, which can be non-linear, from what I will call complete epistemic justification, which is linear and purely deductive. The complete epistemic justification of one's empirical beliefs begins with local justification as its first step.

1. Local Justification. This has two parts. (a) First, one must determine that one believes that p and that one has other beliefs B that are inferentially related to p. (b) One locally justifies the belief p by showing how it is inferable from other beliefs in B, and by showing how it is involved in inference relations to other beliefs in B. Local justification takes place in a context in which the justification of some members of B is taken for granted. Multiple local level justifications can, when combined, produce circular and other closed curves of reasoning.

2. Global Coherence. The coherence of the overall system of beliefs. This stage has three parts. (a) First one must determine one's overall system of beliefs, S. (b) Second, one must determine the coherence of S. (c) Third, on my reconstruction, at this stage one is permitted to make deletions from S to increase coherence. Suppose there is a unique set S' which is the most coherent set obtainable by making deletions from S. Following Leo's suggestion, I refer S' as the harmonization of S.

3. Global Justification or Metajustification. The justification of the overall set of beliefs. I think it is most plausible to think that this set applies not to the original set S, but to its harmonization, S'. At this stage, one determines whether or not S' has sufficient coherence for believing it to be epistemically justified. On BonJour's account, for S' to be epistemically justified, one must be able to show that sets of beliefs with the degree of coherence of S' are likely to be true. So at this stage, one must be able to determine whether sets of beliefs with the coherence of S' are (highly) likely to be true. This requires a metajustificatory argument:

(1) The harmonized set of beliefs S' has degree of coherence c.

(2) Sets of beliefs with degree of coherence of c (or greater) are (highly) likely to be true.

(3) Therefore, S' is (highly) likely to be true.

4. Complete justification of the belief that p. The justification of the particular belief that p, by virtue of its membership in the harmonized system of beliefs, S'. This step simply requires determining that p was not deleted from S when it was harmonized, so that p is a member of S'. If p is a member of S', then p is (highly) likely to be true, and thus, one has completely justified believing that p.