This paper describes experiments carried out at the National Physical Laboratory using a hydrogen maser in comparison with a group of caesium standards which define the NPL time scale. A technique is developed for setting the maser and measuring its frequency with a reproducibility of 1 part in 10¹². Three storage bulbs of different size coated with Teflon FEP-120 and three with Fluon GP1 are used to verify the linear relationship between maser frequency and inverse bulb diameter. The frequency at infinite bulb size is found by extrapolation, the uncertainty in this process being reduced by the use of the two coating materials. In terms of the international time unit, the unperturbed hydrogen hyperfine transition frequency is found to be f_H = 1420405751.7662 ± 0.003 Hz. The effect of contaminating a normally coated bulb with unsintered coating material is investigated. The temperature coefficient of the wall shift for the two materials is investigated over the range 11 to 75° C and it is found that zero wall shift is obtained at 80° C for FEP-120 and at 126 °C for GP1. The results provide experimental verification of the 2nd order Doppler shift within the limits of error of the measurements (± 1 part in 10¹²). Bulbs were measured at intervals over a period of 2 years and there is no evidence of frequency changes exceeding ± 1 part in 10¹². The difficulty of tuning the maser by changing the atomic flux is discussed, and an alternative method is described. It is concluded that the best way of operating the maser as a standard is by continuously resetting the cavity by a servo. A large bulb has a temperature coefficient less than 5 parts in 10¹⁴/°C and the maser could form a definitive standard with only simple control of the room temperature.