When Australia federated in 1901, the Senate (upper house) of the Australian Federal Parliament was envisaged as a house to represent the interests of the constituent states. To investigate the efficacy of this we analyse how what is said in the Senate is affected by the state the senator represents and how this has changed over time. We codify two data sources to create a new dataset of who said what in the Australian Federal Parliament between 1901 and 2018. Using the House of Representatives (lower house) to adjust for common effects, we analyse politician-specific language using a multinomial Bayesian hierarchical model. We explore how the effect of the state that a senator represents changes over time, and hence how the representation of states’ interests has evolved, and how the state-identity of politicians has been subsumed by party-identity in Australia. We find that: the importance of the state-of-origin has decreased over time; and that substantial changes were associated with the world wars. Our findings have implications for how we think about the Senate as either a "states’ house”, or a “party house”.