In first-arrival refraction work, the initial deflection of the first loop of the signal arriving from the shotpoint must be readily recognised against a background of seismic disturbances due to sources other than the explosion. The accuracy of timing a first arrival is determined by this signal-to-noise ratio. It depends primarily on the location of shot and receivers, the size of the charge, and the existing ground unrest at the time of registration. Experiments carried out with these variables kept constant, by recording at the same location from the same shot, show how much the signal-to-noise ratio also depends on the characteristics of the recording equipment used. The best signal-to-noise ratio is certainly not obtained when the transmission curve of the entire system, comprising geophone, amplifier, and galvanometer, peaks at the apparent dominant frequency of the refraction signal. Practical examples show that the signal-to-noise ratio can be improved considerably by using recording systems that transmit a band of frequencies extending many octaves below the observed dominant frequency. The inception of an oscillatory signal was found to be particularly sensitive to the characteristics of a recording system. A seismometer, for example, will transform a starting sine wave with a frequency equal to the natural frequency of the seismometer into a signal with a first loop that is about half as high and half as long as the succeeding loops, the latter moreover being advanced by one-quarter period. This relative constriction of the initial part of a signal is called the "cramping" effect. Such an effect will weaken a refraction first arrival relative to simultaneously arriving later parts of noise signals. This explains why a cramping effect will impair the signal-to-noise ratio. A cramping effect can, of course, be avoided by using a recording system with a flat frequency response. The opposite effect, which can be expected to improve the signal-to-noise ratio, could obviously be achieved by using systems with relatively increased low-frequency response. The practical limit to this improvement would be set by the low-frequency noise that is enhanced by this procedure.