In designing a nuclear reactor fuel pin cluster, particularly one intended for a gas-cooled reactor where the pin surface temperatures can vary considerably over a cross-section, the best available data and methods of calculation are used to determine the pin spacing so as to minimise the maximum surface temperatures for some set of normal operating conditions that may occur in the lifetime of the reactor or for some set of fault conditions that may be postulated to occur. Some features of the design of a cluster or of the operation of the reactor lead to systematic hot-spots at known positions while others lead to systematic hot-spots at random positions. Rather than add these as if they occurred at the same position, as has been done, it is essential to adopt a probability approach in seeking improved reactor performance potential.