There are a number of issues which complicate the interpretation of CMB anisotropy data, some of which we sketch out below.

The microwave sky contains significant emission from our Galaxy and from extragalactic sources. Fortunately, the frequency dependence of these various sources are in general substantially different than the CMB anisotropy signals. The combination of Galactic synchrotron, bremsstrahlung and dust emission reaches a minimum at a wavelength of roughly 3 mm (or about 100 GHz). As one moves to greater angular resolution, the minimum moves to slightly higher frequencies, but becomes more sensitive to unresolved (point-like) sources.

At frequencies around 100 GHz and for portions of the sky away from the Galactic Plane the foregrounds are typically 1 to 10% of the CMB anisotropies. By making observations at multiple frequencies, it is relatively straightforward to separate the various components and determine the CMB signal to the few per cent level. For greater sensitivity it is necessary to improve the separation techniques by adding spatial information and statistical properties of the foregrounds compared to the CMB.

The foregrounds for CMB polarization are expected to follow a similar pattern, but are less well studied, and are intrinsically more complicated. Whether it is possible to achieve sufficient separation to detect B-mode CMB polarization is still an open question. However, for the time being, foreground contamination is not a major issue for CMB experiments.

With increasingly precise measurements of the primary anisotropies, there
is growing theoretical and experimental interest in `secondary
anisotropies.' Effects which happen at *z* << 1000 become
more important as experiments push to higher angular resolution and
sensitivity.

These secondary effects include gravitational lensing, patchy reionization
and the Sunyaev-Zel'dovich (SZ) effect
[41].
This is Compton scattering
( *e*
' *e'*)
of the CMB photons by a hot electron gas, which creates spectral distortions
by transferring energy from the electrons to the photons.
The effect is particularly important for clusters of galaxies, through which
one observes a partially Comptonized spectrum, resulting in a decrement at
radio wavelengths and an increment in the submillimeter. This can be used to
find and study individual clusters and to obtain estimates of
the Hubble constant. There is also the potential to constrain
the equation of state of the Dark Energy through counts of clusters as
a function of redshift
[42].

Although most of the CMB anisotropy information is contained in the power
spectra, there will also be weak signals present in higher-order statistics.
These statistics will measure primordial non-Gaussianity in the
perturbations,
as well as non-linear growth of the fluctuations on small scales and other
secondary effects (plus residual foreground contamination).
Although there are an infinite variety of ways in which the CMB could be
non-Gaussian, there is a generic form to consider for the initial
conditions, where a quadratic contribution to the curvature
perturbations is parameterized through a dimensionless number
*f*_{NL}. This weakly
non-linear component can be constrained through measurements of the
bispectrum or Minkowski functionals for example, and the result from
*WMAP* is -58 < *f*_{NL} < 134
(95% confidence region)
[11].