This paper provides new evidence concerning the interactions between the effects of five year interest rates, leverage levels (the relationship between assets and liabilities), historical losses, Gross National Product (GNP) growth, levels of investment in Research and Development (R&D), income and short-term share returns. The research was carried out using a non-linear model in which neural networks were used to combine different models. The model predicted 31.59% of variation in the returns on shares traded in the AMEX, NASDAQ and NYSE markets during the first five months of all years between 2000 and 2006. The literature suggests that a 32% level of explanation is at the high end, Coad's (2009) literature review of models used to explain company growth finding that linear models explained between 17 and 32% of variability. Like Bode (1998), Tsai (2005), Wang & Chien (2006) and Chien, Wang & Lin (2010) this paper uses artificial neural networks to examine the interaction between R&D and company performance.