We examine the effects of degree of balance between inhibitory and excitatory random synaptic inputs, and of positive correlation between the inputs on the mean and variability of the output of the classical Hodgkin-Huxley (HH) model for squid giant axon, using computer simulation. The mean interspike interval (ISI) and the coefficient of variation of ISI change little as the degree of balance changes, unlike the leaky integrate-and-fire model, frequently used in stochastic network modelling as an approximation to more biophysically based models. Low correlations (up to about 0.1) between 100 excitatory inputs each firing at 100 Hz reduce the mean(ISI) to below a third of its value when the inputs are independent, and CV by a factor of 5 from a near-Poisson range to one associated with regular firing.