Usually this refers to the autocorrelation coefficient. 

Consider any 1D signal with periodicity $\pi$.

Now let's look at the autocorrelation integral:

> $\int_{-\infty}^{\infty}\! f(t)f(t-\tau)\,\mathrm{d}x$

For varying $\tau$, the autocorrelation will have a maximum for $\tau$ equalling $\pi$ and its multiples. Thus autocorrelation can be used to study the periodicity of a signal.


This is often sort of colloquially used to indicate that certain parts of a signal are very similar or even identical.

The analogue for two different signals would be the cross correlation. It can be used to study the similarity of two separate signals.