Temporal characteristics of the web have been analyzed widely in recent years, but graph patterns have served important roles for analyzing the web's structural characteristics. Although temporal characteristics of the web have not been estimated in previous patterns, we specifically examine a novel kind of pattern, time graph patterns, estimating time-series data including the creation times of pages and links. We find useful time graph patterns representing the process by which a topic is discussed extensively during a short period without manual investigations of web graphs. We have also analyzed the patterns and the web pages corresponding to the patterns. Three characteristic pages are contained in the patterns. Additionally, we have succeeded in finding a subgraph matching a mined pattern. We observed that the subgraph corresponds to an extensively discussed topic. © 2010 Springer-Verlag Berlin Heidelberg.