The use in nuclear physics of accelerators giving a pulsed output leads to difficulties when electrical methods are used for detecting the particles produced. The counting losses due to the finite resolving time of the counting system are enhanced by reason of the pulsed nature of the source, and may become considerable unless the counting speed is kept quite low. The present paper contains some calculations on the loss rates to be expected. It is desirable for the resolving time of the counting system, i.e. the dead time following each count, to be much shorter than the pulse length of the accelerator, but since this is itself usually only a few microseconds, this condition is not easy to achieve. If the dead-time is greater than the duration of the accelerator pulse, the calculations are relatively easy but the losses may be high. Calculations for intermediate cases show that the losses can be estimated fairly accurately, and results of practical value can be obtained up to fairly high counting speeds when the dead-time is as high as 40% of the pulse length. The accuracy with which the loss can be calculated will usually be limited by the uncertainty in our knowledge of the shape of the output pulse of the accelerator and of the exact length of the dead-time of the counting arrangement. The attention of the reader is particularly directed to appendix III, where the results of the calculations of this paper will be found summarized. The experimental physicist who wishes to make use of the results without following the detailed analysis may pass directly from the end of section 1 to this appendix, which will also be of value to others for rapid reference.