Here's the text from the book:
The basic problem can be stated in this manner: given a parametric curve Q and a point P, both in the plane, find the point on the curve closest to P (that is, find the parameter value t such that the distance from P to Q(t) is a minimum). Our approach begins with the geometry observation that the line segment (whose length we wish to minimize) from P to Q(t) is perpendicular to the tangent of the curve at Q(t) as shown in Fig. 1. The equation we wish to solve for t is
[Q(t) - P] * Q'(t) = 0.
Luckily, the example given is for a cubic Bezier curve, which is exactly what I need.