Farmland is a fundamental resource for human survival and development. However, farmland fragmentation has become a serious problem, causing ecological damage and low crop production efficiency in many parts of the world. Based on remote sensing and socioeconomic data, we used landscape pattern indices, Morphological Spatial Pattern Analysis (MSPA), and Markov chain models to analyze the temporal and spatial pattern changes in farmland in South Jiangsu Province (the first "Modernization Demonstration Zone" in China) during 1985-2010. Our results demonstrated that the total farmland area decreased by ca. 24% and the farmland pattern became fragmented during 1985-2008: core farmland decreased and islet farmland increased. Additionally, the farmland patch density (PD) increased and three other landscape indices (NLSI, MESH, and COHESION) showed significant decreases. Although the fragmentation rate slowed after 2008, the convergence rate to a stationary farmland distribution became faster, and transitions tended to be less deterministic after 2000. Economic and population growth and policy changes positively contributed to this phenomenon. Therefore, the primary task of farmland protection should be to preserve contiguous farmlands and reduce scattered patches in order to promote farmland connectivity.