For a six month period, serial counts were obtained daily of the blood neutrophils, lymphocytes and monocytes of a patient with cyclic neutropenia. During the last three months eosinophil, basophil, reticulocyte, RBC and platelet counts also were made. Most of these series of blood cell counts showed marked periodicities and these periodicities bore definite phase relationships to one another. Digital bandpass filtering techniques were used to extract an almost periodic signal from the noise fluctuations in each of the observed time series. Phase relationships were determined for the several time series and compared with the results of similar determinations on grey collie dogs with cyclic neutropenia. Comparison of the human and grey collie data demonstrated that the ordering of the lags of the various cell types was the same in the two species. An estimate of the period length of the cycle in both the first half and the second half of the human data was determined using the phase shift over time. This estimate was one and half days shorter for the final three months than for the prior three months of data collection. The particular digital filtering method chosen for this analysis — complex demodulation and remodulation — is well suited for the analysis of periodicities in data where the almost periodic component is changing slowly through time in both amplitude and phase.