Caching is a key technique for improving data retrieval performance of mobile clients. The emergence of robust and reliable peer-to-peer (P2P) communication technologies now brings to reality what we call "cooperating caching" in which mobile clients not only can retrieve data items from mobile support stations, but also can access them from the cache in their neighboring peers, thereby inducing a new dimension for mobile data caching. This work extends a cooperative caching scheme, called COCA, in a pull-based mobile environment. Built upon the COCA framework, we propose a group-based cooperative caching scheme, called GroCoca, in which we define a tightly-coupled group (TCG) as a set of peers that possess similar movement pattern and exhibit similar data affinity. In GroCoca, a centralized incremental clustering algorithm is used to discover all TCGs dynamically, and the MHs in same TCG manage their cached data items cooperatively. In the simulated experiments, GroCoca is shown to reduce the access latency and server request ratio effectively.