Estimates of GDP are sensitive to whether a business expenditure is treated as an investment or an intermediate input. Shifting an expenditure category from intermediate expenditures to investment expenditures increases GDP. While the international guide to measurement (the SNA (93)) recognizes that R&D has certain characteristics that make it more akin to an investment than an intermediate expenditure, it did not recommend that R&D be treated as an investment because of problems in finding a "clear criteria for delineating [R&D] from other activities". This paper examines whether the use of the OECD Frascati definition is adequate for this purpose. It argues that it is too narrow and that attempts to modify the National Accounts would not be well served by its adoption. In particular, it argues that the appropriate concept of R&D that is required for the Accounts should incorporate a broad range of science-based innovation costs and that this broader R&D concept is amenable to measurement. Finally, the paper argues that failing to move in the direction of an expanded definition of R&D capital will have consequences for comparisons of Canadian GDP to that of other countries - in particular, our largest trading partner, the United States. It would provide a biased estimate of Canada's GDP relative to the United States. If all science-based innovation expenditures are to be capitalized, GDP will increase. But it appears that Canada's innovation system is directed more towards non-R&D science-based expenditures than the innovation systems of many other countries. If Canada were to only capitalize the narrow Frascati definition of R&D expenditures and not a broader class of science-based innovation expenditures, we would significantly bias estimates of Canadian GDP relative to those for other countries, such as the United States, whose innovation systems concentrate more on traditional R&D expenditures.