Every idea in this chapter points at one stubborn matrix. The quarter-turn rotation of the plane moves every nonzero vector off its own line, so it has no real eigenvector. Its characteristic polynomial is \lambda^2+1, which has no real roots. The old toolkit of basis, rank, similarity and eigenvalues still applies to it. It needs a larger number system to finish the job.
You can now factor that polynomial over \mathbb C as (\lambda-i)(\lambda+i) and read off the conjugate pair \pm i. You can also tell when a real polynomial must keep a quadratic factor over \mathbb R but splits completely over \mathbb C. Chapter 1 asks a question these numbers raise. Does a vector space need real scalars at all? The answer is no: any number system where you can add, subtract, multiply and divide will do.