Abstract: In process controlling computers, such as switching exchanges, faults may prove to be very expensive. Particularly hard to solve are software defects. If problems occur, it is important that maintenance can be directed at the correct target as soon as possible. A threat is constituted by vandals and hackers. If the option for maintaining from elsewhere is incorporated in computers, the danger for misuse consequently increases. Prior art techniques for detecting memory problems do not necessarily reveal defects occurring at regular intervals caused by the method of calculation. Memory defects will occur at regular intervals e.g., if an address line of a memory circuit becomes faulty. A second drawback of the prior art is that the checksum has to be calculated anew if the contents of just one storage location change. The method of detecting memory problems according to the invention is "regularly irregular" so that it detects in a reliable manner an uncontrolled change in any storage area (MA).