Small phase jitter is the quadrature partner of small amplitude noise.
Say you have a pure carrier and add ordinary white noise, e.g. by putting a resistor in series with the perfect oscillator's output. The resulting RMS phase deviation in some given bandwidth is
= 1/(sqrt(2*CNR))
where CNR is the carrier to noise ratio (i.e. carrier power/noise power in the given bandwidth). The factor of sqrt(2) expresses the fact that the noise and signal are uncorrelated, so that half the noise power winds up in the I phase as amplitude noise, and half winds up in the Q phase as phase noise.
You can derive this from the formula for sums and differences of sines and cosines plus an orthogonality argument--it's quite pretty. It's in my section 13.6 (either edition), but that derivation almost certainly isn't original with me. One very pleasant consequence is that the phase noise statistics are the same as those of the additive noise in the high-CNR limit where the formula applies.
The universality of this formula is why essentially all FM and PM detectors have equivalent performance at high SNR--where the additive model breaks down is low SNR, where FM/PM detection schemes really differ in performance.
Cheers
Phil Hobbs