The effects of daily-read discharges, discharges derived from a single daily observation (say at 9 AM), are examined in the context of flood frequency analysis. These discharges underestimate the true peak discharge because the actual peak is unlikely to occur at the single observation time. The resultant error in discharge estimates can be quite large, up to a factor of 10. Several maximum likelihood methods are proposed to deal with these daily-read, or censored, discharges. Three methods are evaluated: censored as gauged (CAG), which treats daily-read discharges as the true peak, binomial censoring (BC), which uses binomial censoring, and random dependent censoring (RDC), which incorporates the dependence between the the daily-read and true peak discharges. The performance of the methods is evaluated using a Monte Carlo study. The RDC method is found preferable to the CAG method because it provides better performance (in terms of root mean squared error and bias) and is theoretically sound. The BC method performs very poorly and is not recommended.