Circularity is an assumption that was originally introduced for the de
finition of the probability distribution function of complex normal ve
ctors. However, this concept can be extended in various ways for nonno
rmal vectors. The first purpose of this paper is to introduce and comp
are some possible definitions of circularity. From these definitions,
it is also possible to introduce the concept of circular signals and t
o study whether or not the spectral representation of stationary signa
ls introduces circular components. Therefore, the relationships betwee
n circularity and stationarity are analyzed in detail. Finally, the th
eory of linear mean square estimation for complex signals exhibits som
e connections with circularity, and it is shown that without this assu
mption, the estimation theory must be reformulated.