When people use visualizations of conversational archives, they typically reflect on particular events, rather than patterns of activity over time. We explore whether this is a fundamental aspect of how people use data to reflect on the past through pieTime, a visualization we developed that focuses on presenting aggregated behavioral data at timescales from hours to months. It builds on work in conversation visualization and lifelogging by focusing on rhythms rather than details, supporting reflection across media, and using phone logs to complement CMC media. A 15-person evaluation supports findings from prior work about the importance of particular details and storytelling in tools that support reflection, even when the design goals emphasize higherlevel patterns. Still, aggregate patterns provide additional insight into personal behavior, suggesting that systems that integrate both particulars and patterns may be especially valuable, especially when they also help people build and...