When an agent A receives an incoherent turn, he commits to restoring the intersubjectivity in the dialogue; in principle, A has two possibilities: the first one is to change his own view of the interaction; the second one is to persuade his partner B to change his own. The appropriate strategy depends upon the alternative (coherent) interpretation(s) of the interaction that A is able to find out; the existence of such an alternative interpretation is an applicability condition of the action of restoring the intersubjectivity:
1)
If the last (apparently incoherent) turn becomes coherent by changing the
interpretation of some previous turn uttered by himself, then A
can inform B he should correct his interpretation of 
in order to adjust his view of the interaction.
2)
If the last turn becomes coherent by changing the interpretation of some
previous turn 
uttered by B, then A will restructure his own
interpretation of the dialogue.
When the hearer A chooses an alternative interpretation of a turn, he has to reconstruct the interpretation of the whole sequence of turns, in order to identify the partner's dialogue context. In fact, the speaker B who has uttered the problematic turn, should have an interpretation that is coherent from his point of view, although different from A's. Empirical data (see section 5) show that agents succeed in identifying the alternative interpretations of the misunderstood turns, both in cases they have been produced or recognized by them. In fact, repairs are often composed of a part where the recognized wrong interpretation is made explicit (e.g. in turn T4 of Example 1, ``I'm not criticizing ...''). Referring to the classification of repairs in [Schegloff1992], we see that:
1) In third position repair, the speaker of the misunderstood turn stops the diverging dialogue and formulates a repair (a request for a reconstruction of the interpretation), to induce his partner to replace the interpretation of the earlier turns. Schegloff identifies a well defined recurrent schema for formulating a repair, with a few types of realizations and a clear rationality.
2) Instead, fourth position repair looks like a notification that a change of state happened: typically, it is structured as in Example 2 (section 1.2), with a marker like ``Oh" Heritage:84, and a sort of grounding sentence Clark-Schaefer:89, where the agent informs the partner that he has finally understood what the other meant initially.
In Example 1 (Section 1.2), it can be seen that turn T5 ( ``Oh you wanna talk about him''), which follows the request to reconstruct the interpretation of the dialogue, is similar to an acknowledgement of action execution, just as it happens in a fourth turn repair. This similarity suggests that in fourth turn repair, the agent who restructures his own interpretation performs the same action triggered on another agent by a third turn repair. As a fact, both phenomena are characterized by the same underlying goal: the reconstruction of the intersubjectivity. However, the reason for adopting such a goal is different: when a third position repair is performed, the hearer is solicited by the partner's request, which typically also suggests the intended interpretation of the turn ( ``Yeah. I'm not criticizing, I mean we'll just ...''); instead, in fourth position repair, the hearer adopts the goal of recovering from the context misalignment and commits to changing his own interpretation autonomously. In both cases, the speaker notifies the partner that he has restructured his own dialogue context. If, instead, there is no way of realigning the subjective views of the interaction, it is not a problem: unfortunately, intersubjectivity sometimes breaks down; typically, in this case, the failure is notified to the partner.