In this paper the various ways are considered in which a 3-electrode valve should be used in a thermionic voltmeter, and the accumulated experience is described of several years' work by the author in making valve voltmeters for various purposes. No one arrangement can possess every desirable property, and it is shown how to produce a voltmeter to fulfil specified requirements. The effective input resistance of the various arrangements, and the harmonic currents produced by grid current, are the harmonic currents produced by grid current, are considered analytically and experimentally. The effect on range and sensitivity of providing additional batteries is also discussed, extra batteries being found to be an advantage only in very low-range instruments.The effect on the calibration of a change in the valve and in temperature is found to depend on the form of the rectifier used. In general, a thermionic voltmeter cannot read R.M.S. values. The wave-form error of various systems is investigated; with one arrangement the reading is proportional to the peak voltage and in another to the mean. The possibility of a frequency error is discussed; tests show that the error is negligible for frequencies up to 1 000 kilocycles per sec., whilst from indirect tests it would appear that the error is probably negligible for frequencies up to 30 times this value.