Several problems naturally give rise to graphs, and it is often the case that the temporal nature of the data produce time series of graphs. Social networks are a classic example of this, where the ties between actors may change with time, due to changes in the social standing or relationships among the actors. I will describe a simple model which will form the basis for research into graph-valued time series. The model uses the random dot product graph of Scheinerman, which models the graph in terms of attributes of the vertices. I will discuss issues of fitting the parameters, and give some ideas of where the research may lead.