Early reionization changes the pattern of anisotropies expected in the cosmic microwave background. To explore these changes, we derive from first principles the equations governing anisotropies, focusing on the interactions of photons with electrons. Vishniac (1987) claimed that second order terms can be large in a re-ionized Universe, so we derive equations correct to second order in the perturbations. There are many more second order terms than were considered by Vishniac. To understand the basic physics involved, we present a simple analytic approximation to the first order equation. Then turning to the sec- ond order equation, we show that the Vishniac term is indeed the only important one. We also present numerical results for a variety of ionization histories [in a standard cold dark matter Universe] and show quantitatively how the sig- nal in several experiments depends on the ionization history. The most pronounced indication of a re-ionized Universe would be seen in very small scale experiments; the expected signal in the Owens Valley experiment is smaller by a factor of order ten if the last scattering surface is at a redshift $z\simeq100$ as it would be if the Universe were re-ionized very early. On slightly larger scales, the expected signal in a re-ionized Universe {\it is} smaller than it would be with standard recombination, but only by a factor of two or so.

The recent COBE measurement of anisotropies in the cosmic microwave background and the recent South Pole experiment of Gaier {\it et al.} offer an excellent opportunity to probe cosmological theories. We test a class of theories in which the Universe today is flat and matter dominated, and primordial perturbations are adiabatic parameterized by an index $n$. In this class of theories the predicted signal in the South Pole experiment depends not only on $n$, but also on the Hubble constant and the baryon density. For $n=1$ a large region of this parameter space is ruled out, but there is still a window open which satisfies constraints coming from COBE, measurements of the age of the Universe, the South Pole experiment, and big bang nucleosynthesis. Using the central values of the Hubble constant and baryon density favored by nucleosynthesis and age measurements, we find that, even if the COBE normalization drops by $1\sigma$, $n > 1.2$ is ruled out.