Newton's method basins of attraction by graphical iteration

From the form Nf(x) = x - (f(x)/f '(x)) we see that roots of f(x) are fixed points of Nf(x):
 
    If f(x*) = 0 and f'(x*) ≠ 0, then it is clear that Nf(x*) = x*.
    Even if f '(x*) = 0, for fairly general functions f(x) (any function represented by a Taylor series will work), the function Nf(x) can be defined on x*, and for this extended function Nf(x*) = x*.
Moreover, simple calculations show that each root of f(x) is a stable fixed point for Nf(x).
 
Finding the basins of attraction of the roots of f(x) is relatively straightforward using graphical iteration of Nf(x).
This is most easily seen through an example. We take f(x) = (x+1)x(x-1), a function whose roots are -1, 0, and 1.
On the left (a) we see the graph of Nf(x); the vertical dashed lines are the points at which f '(x) = 0.
Certainly, we see that iteration from an initial point to the right of the right vertical dashed line converges to x = 1, and iteration from an initial point to the left of the left vertical dashed line converges to -1. Are these the entire basins of attraction of +1 and -1?
 
The answer is "No." We can see this by reversing graphical iteration.
Denote by P and Q the intersections of these vertical dashed lines with the diagonal.
Now do reverse graphical iteration from P and Q. That is, go horizontally to the graph, vertically to the diagonal. See the right picture (b).
This produces sequences of intervals converging to +1 (indicated by the bars above the x-axis), and to -1 (bars below the x-axis).
Note these reverse graphical iteration paths appear to converge. In fact, they converge to a 2-cycle for Nf(x). Iteration from points within the 2-cycle converges to x = 0.

Return to real Newton's method.