The objective of this study was to provide a more complete understanding of the changing pace and nature of production and productivity growth in South African agriculture during the 20th century and the associated changes in research and development (R&D) investments and institutions that affect agricultural input, output and productivity performance. A completely new panel of data was constructed to track investments in agricultural R&D and scientific capacity that took account of the numerous structural and organizational changes that shaped public R&D since 1910. The national agricultural production accounts were also revised to address the legacy of South Africa's history of racial segregation and a multitude of problems that arose in the official time series data due to changes in the underlying statistical methods and procedures. With these new output and input data in hand the evolution of production agriculture over the past century was quantified and described. The modern indexing methodologies deployed in this study, in conjunction with new primary price and quantity data yielded new insights into the economic evolution of South African agriculture over the past century. This study analyses the changing historic patterns of public sector investment in the agricultural sector and identifies the phases in policy evolution against the trends in aggregate spending on agriculture, farmer support and R&D. Following an initial phase of scientific capacity building, the R&D system developed a measure of synergy in its activities as evident in the spending patterns of the national and regional institutes from 1926 to 1971. The concordance of policy and institutional changes with R&D investment, output and productivity trends in the funding of the various research entities came to an end in 1980, and overall public investment in agricultural R&D has stalled since 1978. Growth patterns in multi-factor productivity estimated in this study substantially differ from earlier studies, especially in terms of magnitude, and present different results on the estimates of the growth in agricultural output that is attributable to productivity growth. It was found that not only did the earlier methods yield indexes that overstate growth patterns — thus suffering from aggregation bias in their index numbers — but trended more erratically and in poor concordance to the timing of policy changes.