Lets say I have a SLR with correlation coefficient (r) = -0.81 then the coefficient of determination (r^2) will be around 0.661. From what I know, this would mean that will the while the variables have a strong negative correlation the independent variable is only able to account for 66.1% of the variation in the dependent variable. My question then based on this is, what does this suggest? Is the relationship between the two variables non-linear or I need to add more independent variables?