The Gaussian noise burst model and the domain model for imperfectly mode-locked pulses are examined through comparison with two types of two-photon correlation experiments. These models are known to agree with autocorrelation measurements using nonlinear optical methods. The Gaussian noise burst model does not give correct results for the pulse-cycleaveraged, normalized second-order correlation function at zero time-delay, and the domain model fails to explain the two-component dip in coincidence count rate curves obtained with the interferometer configuration.