Cornering is a change in the direction of motion of a car. In order to change the direction of motion, we must change the direction in which the car is pointing. To do that, we must rotate or yaw the car. However, the car will resist yawing because the various parts of the car will resist changing their states of motion. Let's say we are cornering to the right, hence yawing clockwise. The suspension parts and frame and cables and engine etc. etc. in the front part of the car will resist veering to the right off their prior straight-line course and the suspension parts and frame and differential and gas tank etc. etc. in the rear will resist veering to the left off their prior straight-line course. From this observation, we can 'package' the inertial resistance to yawing of any car into a convenient quantity, the PMI. What follows is a simplified, two dimensional analysis. The full, three-dimensional case is conceptually similar though more complicated mathematically.

It turns out that the general motion of any large object can be broken up into the motion of the centre of mass, treated as a small particle, and the rotation of the object about its centre of mass. This means that to do dynamical calculations that account for cornering, we must apply Newton's Second Law, F = ma, twice. First, we apply the law to all masses in the car taken as an aggregate with their positions measured with respect to a fixed point on the ground. Second, we apply the law individually to the massive parts of the car with their positions measured from the CM in the car while it moves.

Let's make a list of all the N parts in the car. Let the variable i run over all the limits in the list; let the masses of the parts mi, their positions on the X axis of the ground coordinate grid be xi and their positions on the Y axis of the ground coordinate grid be yi. We summarise the position information with vector notation, writing a bold character, ri, for the position of the i-th part. Vector notation saves us from having to write two (or three) sets of equations, one for each coordinate direction on the grid. For many purposes, a vector can be treated like a number in symbolic arithmetic. We must break a vector equation apart into its constituent component equations when it's time to do number-crunching.