In this paper we have explored some of the ways in which the structure of the lifeworld supports agents' cognition, and we have suggested how this analysis might be expanded to cover a wider range of phenomena. Much work obviously remains to be done. Perhaps the most significant part of this work concerns a fundamental assumption of lifeworld analysis: that people use objects in customary ways. This is a plausible enough first approximation, but it is not always true. Faced with a difficulty that goes beyond the capacities of the usual practices and the artifacts that are readily available, people frequently improvise. The handle of a spoon might be used to pry open a lid, a pen might be used to fish acorns out of an exhaust duct, a book might be used to provide backing for a sheet of paper one is writing on, or a protruding section of a car's bumper might be bent straight by deliberately driving the car into a concrete wall. In these cases the underlying physical affordances of an object ``show through'' beyond their ready-to-hand appropriation in routine patterns of interaction. These underlying affordances can also show through in situations of breakdown, for example when a tool breaks or proves inadequate to a job. In such cases, people confer improvised meanings upon artifacts. Such phenomena are particularly important in conversation, in which each utterance is interpreted in the context created by previous utterances, while simultaneously helping to create the context for interpretation of successive utterances as well [13, 5]. The point is not that the lifeworld does not exist, but rather that it is something actively created as well as something adapted to through socialization. One challenge for future research is to learn how computational methods might help in modeling such phenomena--and how such phenomena might help us to rethink basic ideas about computation.