The mesencephalic reticular formation (MRF) neurons are regarded as contributing to the activation of the celebral cortex. In this paper, the statistical features of single neuronal activities in MRF of cat during dream sleep are investigated; the neuronal spike train exhibits 1/f fluctuations. Counting statistics is applied to the neuronal spike train giving rise to a variance/mean curve which follows a tμ-law. For an interpretation of these findings, the clustering Poisson process is applied which not only gives rise to a tμ-law but also suggests a generation mechanism. The MRF neuronal activities are closely fitted by the clustering Poisson process and the underlying statistical parameters can be estimated. These findings strongly suggest that neuronal activities can be interpreted as superposition of randomly occuring clusters ( = bursts of spikes).