Abstract: In this paper, a view in retrospect is made on some highlights in the early progress of numerical weather prediction(NWP), trying to address their implications in methodological perspective. NWP as a branch of atmospheric sciences, is a subject of interdisciplinary nature and immediate practical use. In the course of its development, the raising of a new idea or the advent of a new research area in NWP, not seldom, stems from its integration with other disciplines, let alone the demand and inspiration of weather prediction practice. From this viewpoint, two streams of researches have been traced with respective headings of “from the ‘dream’ of Richardson to the recurrence of primitive equations model” and “from the initiation of using historical data in NWP to the formulation of 4-D data assimilation” in the text.