Short-term correlations in activity have been widely used as evidence to connect brainstem units with postganglionic sympathetic nerves. These may be detected by spike-triggered averaging, cross correlation of coherence analysis. The specificity of this type of evidence has been investigated by cross-correlating the activity of identified cutaneous vasoconstrictor postganglionic fibres with that of medullary premotor neurones of like and of unlike functional type, as determined by physiological testing (preoptic warming), in anaesthetised cats. Single medullary premotor neurones of both types were recorded from the subretrofacial nucleus: they were identified by their berosensitivity and, in most cases, their spinally projecting axons. By the test criteria chosen, the correlation method gave both false-positive and false-negative results as commonly as it gave correct ones. We conclude that it is not a reliable way to determine brainstem-postganglionic connectivity.