Nonsense results when constraining a parameter to be greater than a negative value.

There is a bug in Prism 5.00 (windows) that was fixed in 5.01. If you constrain a parameter in nonlinear regression to be greater than some negative value, in fact Prism constrains it to be greater than the absolute value of that number. So if you set a constraint that K be greater than -10, Prism actually constrains K to be greater than 10.

If you constrain a parameter to a constant value, it works fine -- both when that value is positive and negative.  Constraining to be less than a value also always works, with both positive and negative values.