Since polarization directly probes the last scattering epoch the first thing
we learn is when that occurred, i.e. what fraction of photons last scattered
at 
when the universe recombined, and what fraction rescattered
when the intergalactic medium reionized at 
.
Since rescattering erases fluctuations below the horizon scale
and regenerates them only weakly ([Efstathiou] 1988),
we already know from the
reported excess (over COBE) of sub-degree scale anisotropy that the
optical depth during the reionized epoch was 
and hence
It is thus likely that our universe has the interesting property that both the recombination and reionization epoch are observable in the temperature and polarization spectrum.
Unfortunately for the temperature spectrum, at these low optical depths
the main effect of reionization is an erasure of the primary anisotropies
(from recombination) as 
. This occurs below the horizon at last
scattering, since only on these scales has there been sufficient time to
convert the originally isotropic temperature fluctuations into anisotropies.
The uniform reduction of power at small scales has the same effect as a
change in the overall normalization.
For 
-20 the difference in the power spectrum is confined
to large angles ( 
).
Here the observations are limited by ``cosmic variance'': the fact that we
only have one sample of the sky and hence only 
samples of any given
multipole.
Cosmic variance is the dominant source of uncertainty on the low- 
temperature spectrum in Fig. 14.
The same is not true for the polarization. As we have seen the polarization spectrum is very sensitive to the epoch of last scattering. More specifically, the location of its peak depends on the horizon size at last scattering and its height depends on the duration of last scattering ([Efstathiou] 1988). This signature is not cosmic variance limited until quite late reionization, though the combination of low optical depth and partial polarization will make it difficult to measure in practice (see Fig. 14 and Table 2). Optical depths of a few percent are potentially observable from the MAP satellite ([Zaldarriaga et al.] 1997) and of order unity from the POLAR experiment ([Keating et al.] 1997).
On the other hand, these considerations imply that the interesting polarization signatures from recombination will not be completely obscured by reionization. We turn to these now.