Synchronized neuronal activity is an efficient way to propagate activity patterns spontaneously in a neuronal network. A cultured neuronal network commonly shows synchronized network bursts, the number of reproducible repertoires of which may represent inner information capacity of the network. However, previous studies mainly focused on the statistical features of network bursts and overlooked the similarity of sequential spatial activity patterns. We cultured neurons derived from the cortex of Wistar rat fetus on a high spatial-resolution microelectrode array, consisting of 4,096 electrodes in 2.67 mm Χ 2.67mm area at 42 μm resolution, and investigated spontaneously occurring network bursts at the early stage of development. The non-negative matrix factorization decomposed recorded spike data into the bases of spatial activity pattern and the time series of coefficients. We then evaluated the coefficient transition in the network burst periods in terms of reproducibility, and found similarity in spatial propagation patterns. Thus, our results suggest that the activity pattern was reproduced in the network bursts at the early developmental stage.