Summary In this paper we report first results of our study on network characteristics of a reference-based, bibliographically coupled (BC) publication network structure. We find that this network of clustered publications shows different statistical properties depending on the age of the references used for building the network. A remarkable finding is that only the network based on all references within publications is characterized by a degree distribution with a power-law dependence. This structure, which is typical for scale-free networks, disappears when selecting references of a specific age for the clustering process. Changing the publication network as a function of reference age, allows 'tuning through the episodic memory' of the nodes of the network. We find that the older the references, the more the network tends to change its structure towards a more exponential degree distribution.