I will have to look at what the SIGN model entails. Is this what is also refers to as a +/- signed bit model ? So all you know is the change of sign? This approach is often used in qualitative reasoning models.
So then one can come up with a correlation of unity if all (+)*(+) or (-)*(-) give a +1 value and 0 if random. The amplitude is out of the picture so then one doesn't have to worry about slight non-stationary amplitude variations, caused by noise, etc.
I was thinking about what the modulated sinusoid means.
Say that a function is described as:
$ f(t) = sin(wt+sin(vt)) $
the first derivative of this
$ f'(t) = (w+v cos(vt)) cos(wt+sin(vt)) = (w+v cos(vt)) \cdot f(t+\pi/2) $
which indicates that a time-modulated interaction may be generating the resulting modulation, ala a Mathieu or [Hill differential equation](http://en.wikipedia.org/wiki/Hill_differential_equation).
This gets back to my original premise that these perturbed differential formulations may be very common in understanding climate to first-order.