Time resolved correlation (TRC) is a recently introduced light scattering technique that allows one to detect and quantify dynamic heterogeneities. The technique is based on the analysis of the temporal evolution of the speckle pattern generated by the light scattered by a sample, which is quantified by cI(t, tau), the degree of correlation between speckle images recorded at time t and t + tau. Heterogeneous dynamics results in significant fluctuations of cI(t,tau) with time t. We describe how to optimize TRC measurements and how to detect and avoid possible artifacts. The statistical properties of the fluctuations of cI are analyzed by studying their variance, probability distribution function, and time autocorrelation function. We show that these quantities are affected by a noise contribution due to the finite number N of detected speckles. We propose and demonstrate a method to correct for the noise contribution, based on a N--> infinity extrapolation scheme. Examples from both homogeneous and heterogeneous dynamics are provided. Connections with recent numerical and analytical works on heterogeneous glassy dynamics are briefly discussed.