The study of the cosmic microwave background (CMB) radiation has had a long history. Three aspects of the CMB might be considered: its existence, its spectrum, and its anisotropies. By firmly establishing that the Universe expanded from an initially hot, dense state, the existence of the CMB underpins our entire cosmological framework. It has been recognized from the beginning as one of the pillars of the hot big bang cosmologies. The spectrum of the CMB is the most precise blackbody spectrum in nature, from which many inferences can be made. Although this discovery is less than a decade old, its impact on models of the early Universe been discussed extensively elsewhere, (e.g. [Nordberg & Smoot 1998]). In this paper we would like to consider the anisotropies in the CMB, the small fluctuations imprinted on the sky by the progenitors of the large-scale structure seen in the distribution of galaxies today.
In the roughly seven years since the COBE team announced the first detection of anisotropies in the CMB ([Smoot et al. 1992]), more than a dozen groups have reported detections, covering the full range of frequencies and a wide range of angular scales (see [Smoot & Scott 1998, Bennett, Turner & White 1997]). Due in large part to a dramatic increase in detector sensitivity, mapping the CMB anisotropy has become almost routine. Our confidence in the results has grown as multiple observations by the same teams over a period of years, and then later by different experiments at different frequencies and sites, reproduced the same features on the sky and confirmed their black body nature.
Over the same period much progress has been made in data analysis techniques and in the theoretical interpretation of CMB data. Better physical understanding of the anisotropy generation has lead to faster algorithms for its computation (e.g. [Seljak & Zaldarriaga 1996, Hu et al. 1998]) applicable to an impressively wide range of theories. The high precision calculations and accurate measurements of the anisotropy have spawned numerous ideas in data analysis, with a full likelihood analysis of mega-pixel CMB maps now within reach ([Oh, Spergel & Hinshaw 1999]).
However, since much of the progress has been incremental, it is not always
obvious just how far we have really advanced. It therefore seems appropriate
to take stock and ask the question: