Nonlinear regression assumes that the average distance of points from the curve is the same for all parts of the curve (homoscedasticity) or that you have chosen an appropriate option on the weights tab to correct for this problem.

Prism 7 can test for homoscedasticity or appropriate weighting. To do this, Prism 7 computes nonparametric correlation between the absolute values of the weighted residuals and the Y value of the curve. A low P value from this test means that the weighted residual is correlated with the Y value of the curve, and this suggests  the weighting choice was not ideal.