California responded to concerns about overfishing in the 1990s by implementing a network of marine protected areas (MPAs) through two science-based decision-making processes. The first process focused on the Channel Islands, and the second addressed California's entire coastline, pursuant to the state's Marine Life Protection Act (MLPA). We review the interaction between science and policy in both processes, and lessons learned. For the Channel Islands, scientists controversially recommended setting aside 30-50% of coastline to protect marine ecosystems. For the MLPA, MPAs were intended to be ecologically connected in a network, so design guidelines included minimum size and maximum spacing of MPAs (based roughly on fish movement rates), an approach that also implicitly specified a minimum fraction of the coastline to be protected. As MPA science developed during the California processes, spatial population models were constructed to quantify how MPAs were affected by adult fish movement and larval dispersal, i.e., how population persistence within MPA networks depended on fishing outside the MPAs, and how fishery yields could either increase or decrease with MPA implementation, depending on fishery management. These newer quantitative methods added to, but did not supplant, the initial rule-of-thumb guidelines. In the future, similar spatial population models will allow more comprehensive evaluation of the integrated effects of MPAs and conventional fisheries management. By 2011, California had implemented 132 MPAs covering more than 15% of its coastline, and now stands on the threshold of the most challenging step in this effort: monitoring and adaptive management to ensure ecosystem sustainability.