The spontaneous and evoked activities of individual pairs of single units were recorded simultaneously with the same microelectrode in the medial geniculate body (MGB) of ketamine-xylazine-anaesthetised guinea-pigs. Cross-correlograms (CCGs) of spike train pairs were computed and divided on the basis of correlation peak shape into four classes [a unilateral narrow (UN) peak, a centrally positioned wide (CW) peak, a complex peak and no significant peak] interpreted in terms of the functional connection between neighbouring neurones. The shift predictor procedure was applied with the aim of removing the effect of the stimulus on the final CCG shape. The occurrence of correlation peak types and the distribution of correlation coefficients were found to be similar for the spontaneous activity during silent periods following acoustical stimulation and for the long-lasting recording of spontaneous activity. CCGs in 38% of pairs computed during silent interstimulus intervals contained a UN peak, suggesting a monosynaptic excitatory connection. Almost 20% of all pairs expressed a CCG shape typical for a common input, i.e. a CW peak. In 5% of cases multiple, so-called complex peaks, were found. About 20% of the CCGs contained no significant correlation peak in the interstimulus period, which is typical for a very weak or absent functional connection between recorded neurones. No inhibitory interaction (groove in the CCGs) between recorded pairs was observed. The distribution of correlation peak shapes was similar when calculated during acoustical stimulation and during silent interstimulus intervals. CCGs computed during presentation of four acoustical stimuli (pure tone bursts, noise bursts, natural call whistle and artificially inverted whistle) showed most frequently a UN peak (28-37%) followed by CCGs with no significant peak (18-28%) and with a UN/CW peak (14-23%). On average, the occurrence of UN peaks tended to be less frequent during stimulus presentation than in silent conditions, but the difference was not statistically significant. The most frequent occurrence of clear UN peaks was found in the medial part of the MGB (from 52-64% of pairs depending on the type of acoustical stimulus), while the least was observed in the ventral part of the MGB (12-22%). In contrast, CW peaks were most frequently expressed in pairs located in the ventral part of the MGB (18-33%), while neuronal pairs in the medial part revealed a very low occurrence of CW peaks (0-7%). The occurrence of independently firing neurones was lowest in the medial part of the MGB (8-20% of pairs) in comparison with the ventral (31-39%) and dorsal (12-41%) parts. In 20% of pairs acoustical stimulation produced a change in the type of correlation peak present during spontaneous activity. Most frequently, a CW peak (shared input) changed to a flat CCG, which represents independently firing neurones. In some pairs higher connection strengths (as expressed by the value of the correlation coefficient) were found for silent interstimulus intervals than for acoustical stimulation. This suggests that in the MGB the stimulus may desynchronise the spontaneous activity of simultaneously firing units in neuronal pairs.