If the underlying trials are positively dependent (meaning that a success, say, is more likely followed by another success), would you expect the variance of the total number of successes to be larger or smaller than the binomial expression np(1 – p)?