Description
Consider this pseudocode:

class A {
    int f;
}

x = new A();
x.f = 5;

With a field sensitive analysis we can correctly separate not only the values of x and x.f, but also their types (A* and int). Whenever we use a field insensitive analysis, x.f will point to the same location as x, meaning that the type inference will infer the types of x.f as A* since it alredy has a mapping from x to A* in memory when the field access is reached.