This article discusses a communicative phenomenon that is relatively less studied: getting stuck in an aphasic conversation. Although aphasia as a medical and linguistic condition has been widely examined, the more social and participatory aspects of the symptom are not so well-known. Aphasia forms a threat to the emergence of a shared understanding, as well as to the experience of being in the shared, i.e., in the intersubjective, social world. In our analysis, we closely explore how understanding is constructed in the sequential organization of conversation. We use two data corpora when analysing the halting interaction. In our data, we detected two kinds of interactive halts that emerged in connection with aphasic word searching. First, 'real halts' were caused by the aphasic person's inability to find correct words and the co-participants were also not able to resolve the problem. Second, 'exam halts' occurred when the co-participant did not provide the missing words despite knowing what the aphasic speaker was trying to say. We discuss how this halting phenomenon is linked with the notions of intersubjectivity and face-work and conclude that real halts are more directly caused by the aphasic condition, whereas exam halts reflect the spousal relationship in the form of face-work.