We argue that the notion of scientific information integration (SII) should be broadened to include the emerging paradigm of scientific workflows. To the domain scientist, SII often means an "end-to-end" process, consisting of data acquisition, transformation, analysis, visualization, and other steps. We argue that SII requires a new paradigm that combines a data-oriented approach with a suitable process-oriented workflow modeling approach. Our experiences in a number of collaborative projects indicate that scientists have a real need for this extended kind of information integration. Towards this end, we introduce a number of concepts such as models of computation and provenance, actor- and flow-oriented programming, higher-order components, adapters, and hybrid types and then outline a particular combination of these, yielding a promising new synthesis of process and data modeling called Collection-Oriented Modeling and Design.